7.01.2017

CAUTION!!!! BANDING!!!! OMG. OMG.

A shot on stage with LED stage lights at Zach Theatre. Sony A7Rii. 70-200mm f4.0. 1/400th shutter. While banding was not apparent to human eyes watching the play it was horrifically obvious to the electronic shutter in my Sony A7Rii.

You really don't have to go far to find banding in just about any camera that uses an electronic shutter. The camera shutters scan from top to bottom. If the light source creating the image is not a constant source there is a probability that you'll see some banding at one point or another. It's part of every alternating current light source. The only light sources that are truly constant are direct current powered light sources. In days of yore even entry level photographers knew that shooting under (badly ballasted) fluorescent lights would cause banding unless you used a shutter speed long enough to allow the band to travel all the way down each frame before the shutter closed. 1/60th was possible but 1/30th was safe. Going into shutter speeds above 1/125th of second could almost guarantee banding and I have countless examples of this fluorescent light banding in conjunction with Nikon D700's, Nikon D750's and any number of Canons. All cameras without electronic shutters. 

For the last two days the folks at DPReview have been running a "Have you stopped beating your wife yet?" faux scientific article trying to explain why Sony's new a9 camera showed banding when shooting LED driven screens at a sports event. It is in response to
a lone banding sighting by a blogger on the web. The short answer is that LED instruments not made for movie filming and photography can have obvious scan differences over time. Most instruments (LED projectors, LED screens, etc.) have very fast on/off times and can be safely used for most professional video applications where the shutter speeds are in the range of 1/50th, 1/60th, etc. Since they switch on and off at high multiples of the line current and are also dependent on the frame rate/refresh rate (never mentioned!!!!!) of the programming on the screen, at some point banding is inevitable. The faster the shutter speed the finer the banding. The pseudo science of the web is quick to blame the shutter mechanism of the a9 while the actual LED lights involved yield an equally plausible answer that the refresh rate of the video programming being used on the instrument, and the cycle of the LED is to blame.

In the photographic examples above and below the theater I was shooting in used lights that are not optimized for movie, video or still photography capture at faster shutter speeds. They are more economical than lights optimized for those uses but they are more than adequate to cover persistence of human vision and seem interchangeable (as far as the audiences are concerned) with traditional tungsten floods. Tungsten lights also turn on and off but the filament temperatures don't drop enough in the "off" cycles for the fluctuations to be visible. The filament continues to glow even during a "no current" phase. There is much less obvious decay but one will only see it if shooting at crazy high shutter speeds. 

The implication of the article is that this problem just came to light in Sony's newest camera (they do mention that it can occur in mech shutters but only as a small side note). As you can see from the examples here banding is a known possible hazard when using any light source that cycles with line current or refreshes at a specific rate. It affects every electronic shutter in exactly the same way. No camera using an electronic shutter is immune. In fact, no camera with a focal plane shutter is completely immune either. You want to catch a shot without the banding effect? Shoot with a mechanical leaf shutter. Seems to be highly resistant to all these banding effects. 

But let's not get carried away and convict Sony of creating a banding monster since it is no more or less bandable than any of it's close competitors who offer electronic shutters. 

A shot on stage with LED stage lights at Zach Theatre. Sony A7Rii. 70-200mm f4.0. Front spot is not LED on this image. See the difference the between image at top and this one. But notice the banding on the background and compare it with the background of the image below.

 A shot on stage with LED stage lights at Zach Theatre. Sony RX10iii. shutter 1/200th. This image is using the same rear projectors as the image just above but the mechanical leaf shutter does not scan, the entire image is captured with the shutter fully open. Hence, no scan lines on the background. 


As just above. 

Back to the A7Rii and notice the repeating, strong red lines in the background. They were not visible to the eye when viewing the play. They are a result of the shutter scanning and the frame refresh rate of the projectors. You can't always escape it by using faster shutter speeds but longer shutter speeds will blend the lines together in the final frame much as the human eye does.

Front "continuous" light spot means no banding in the front plane of the photo (just above).
The banding was caused by the rear projection. 

But, I am just a photographer so if you have more information or a better handle on this than me or DPReview please state your credentials and then wade right into the comments....

10 comments:

Michael Matthews said...

Remember all those feature films and TV shows from years past when CRT computer montors in the background were seen as continuously scrolling, rolling upward? You'd think that with all that production money flying around someone would have thought to change the monitors' refresh rate or the camera's shutter speed to make the problem go away. TV news crews frequently ran into the same problem shooting in real world settings -- businesses with computer screens visible, interviews in homes with a TV set on in the background, etc.. That finally got fixed with the introduction of ENG cameras with "clear scan", a dial-it-in setting which let the camera operator vary the shutter speed by increments of .2 Hz on the fly until the rolling screens and scrolling stripes vanished.

Kirk Tuck said...

Absolutely!!!! That DPReview wasted so much energy trying to re-invent a problem is just astounding. How many times does the wheel need to get invented before we can move on?

David said...

Excellent, write up. This actually explains the problem, well done kirk.
The problem with the dpreview article is Rishi.
First they need page clicks. This is obvious as the switch to more petapixel type news reporting. This I am ok with, but can lead to sensational news headlines.
But then there is Rishi, with a Ph.D in Biophysics, whom in jumping into water as a known expert, that I don't think he should be in. I too have a Ph.D in Biophysics.
With this dpreview shift. Rishi is becoming the "Dr. Phil" of tech.
Your description is more helpful, explains mixed lighting and is also not as biased. But will it get as many page clicks and comments? That is the issue.
I am now waiting for the "have you stoped betting your wife" articles to test and compare body durability. Couldn't we hammer nails with our Olympus E3 bodies. I still have no fear handing that camera over to my kids to take images.

Kirk Tuck said...

Thanks David. I'm not a Phd. in Biophysics but I did write the definitive book on LED lighting for Photographers and I do have more than a basic understanding about LEDs, non-continuous light sources, basic electricity and common sense. Plus, I probably have 30+ more years of experience in shooting photography under all sorts of lighting conditions. Banding caused by fluctuations in lights is hardly rocket science. It's sad that a website like DPReview would try to make it seem so. Experimental results are only valuable if they are repeatable. At least that's what all the scientists here at The Visual Science Lab tell me. But I speak to them infrequently because they are busy trying to upgrade the CERN supercollider to get premium cable TV....

Rishi seems busy trying to convince people that DPReview is akin to the "Big Bang Theory". He may be their Sheldon Cooper.

Maybe some more real scientists like you will chime in here as well.

Noons said...

Thank the Gods for folks like you, Kirk.
The field of photography is sooooo full of pretend geniuses, I often despair of getting any useful info...

Bassman said...

Correct me if I'm wrong, but the same problem appears with mechanical shutters only if shooting faster than the flash sync speed. That is because the mechanical FP shutter is never fully open, but forms a slit that travels across the sensor (or film).

Wally said...

OMG OMG OMG an article about a technical issue in Photography that makes sense, is written in plain English, and can't be explained away by fanboyism or reviewer stupidity/propaganda/need to generate controversy to make ratings. OMG OMG OMGMG

Scott said...

I've always heard that Austin is a great place to see bands. :)

Unknown said...

I'm not sure what we've done to invoke your ire on this one.

A popular blogger posted a video suggesting the a9 had a problem - and one that there'd been concerns about, when it was clear that the a9 is a primarily e-shutter camera. We felt we could add some facts (and detail) to the discussion.

Rishi's piece is a bit wordy, but his fundamental points:

1) That the scanning nature of all rolling shutters (including focal plane shutters) is interacting with the flicker/refresh of an unusual light source (an LED advertising hoarding), is the underlying cause.
2) That the a9's 12-line readout is giving that banding a hard edge

Don't seem unreasonable.

Perhaps we overestimated how widespread the chatter and concerns were, but I don't think we were wrong to publish a response.

Given we link back to an article I wrote about shutter behaviour and effects some weeks before, it's clear that (12-line readout contribution aside), we didn't think we had stumbled on anything revolutionary. Similarly, we're getting as many people convinced we're trying to hush something up as we are claiming we're making something out of nothing, so it's hard to know what we've done wrong.

Kirk Tuck said...
This comment has been removed by the author.