Can someone explain in simple terms what's causing this?

[Unframed Dave]

The striping?

I have read several explanations, but as someone with few cells left to spare, I'm still struggling to understand. Personally, I find it easier to prevent something happening again, if I can understand it, rather than learn rote the corrections required.

I did pick the wrong camera on this occasion, I should definitely have used the A1 over the A7CR. Unfortunately, the environment wasn't the best for changing lenses.

• ILCE-7CR
• FE 24-70mm F2.8 GM II
• 24.0 mm
• ƒ/4
• 1/1000 sec
• ISO 5000

 

[Alan Clark]

This article seems to explain it. You can fix it in two ways: use the mechanical shutter instead of the electronic shutter, or use the Anti-flicker shoot setting on the camera.

https://www.cacei.ro/how-to-fix-banding-flickering-for-sony-a7-and-a1-series-quick-and-easy/

 

[brphotographycanada]

The article Alan links has an in depth explanation. I will attempt a "TL;DR" version.

The electronic shutter "scans" through the sensor line by line, pixel by pixel. This takes a certain amount of time. The indoor lighting being alternating current (AC), basically turns on-off rapidly. As the sensor scans through, the lights turn on and off giving uneven exposure and the bands you see.

The mechanical shutter physically opens and closes to take an exposure. The "smarts" in the camera detect the flicker and sync the shutter to operate "in time" with the artificial lights giving a more even exposure and therefore no banding.

With fast moving subjects like you are seeing, there is another phenomena where objects stretch under the electronic shutter. You can see a bit of this with the riders right arm (maybe it's just a baggy shirt lol). Again the mechanical shutter solves this.

 

[Thad E Ginathom]

Silent shutter... such a great idea until one actually uses it, indoors, in artificial light --- and it fails in the very situation it's most needed in.

It wouldn't if it wasn't for LEDs? Yes, but what lighting isn't LED these days!

 

[Unframed Dave]

Thanks all, I now understand.

 

[Unframed Dave]

OK, so having slept on it, and in between being a ninja turtle, spending a while on a desert island with Kelly Brook, and setting up a smallholding with Felicity Kendal in seventies suburbia, my thoughts turned back to banding. To be honest, banding wasn't my favourite. Anyway.....

First up, this electronic shutter thing is mind blowing. I can be shooting an owl at 30 fps and the sensor is "reading" light pixel by pixel, line by line in a 3000th of a second (or less). We live in remarkable times.

Secondly, I thought I understood, but now, I'm not convinced that I do.

If the statement above is a correct understanding, then I'm missing something.

UK alternating current switches every 50th of a second, and if my schoolboy physics and rapidly failing memory is holding up, it doesn't switch back and forth, it's sinusoidal.

So my questions are:

a. the shutter speed above is 1/500, why is it registering differences in light occurring at 1/50 of a second.

b. why are the bands distinct? If the light change isn't on/off, rather gradations in between on and off, surely the bands would follow the same pattern.

Yours

Still bewildered from Norfolk

PS, not arguing, merely trying to fill in the gap in my understanding.

 

[Uncle Kevriano]

The difference in E Shutter on the CR and the A1 is also one of the reasons you may not have seen it/noticed it before, because the CR is not a stacked sensor, where as the A1 is. It's another reason I don't use it on the RiV

 

[Alan Clark]

a. the shutter speed above is 1/500, why is it registering differences in light occurring at 1/50 of a second.

I don't fully understand the patterning myself, but here are a few points:-

1. The voltage goes from +220 to -220 to +220 every 1/50s, so peak voltage, and therefore the flicker rate will be 100Hz, not 50.

2. The shutter speed is 1/1000, not 1/500, according to the metadata.

3. The shutter speed is the amount of time that a pixel is exposed to the light, but not all pixels are exposed at the same time (except with a global shutter, which you do not have). The time needed to collect all of the image is longer than this.

 

[Unframed Dave]

1. Gotcha

2. My mistake, but that aids my confusion

3. Now, I need to get my head round that, but I think I'm following it.

Thanks for the input.

 

[Angus Cameron]

There is some good information here without the explanation being overly complex

https://www.sylvania-lighting.com/media/5031/sylvania_white_paper_understanding_led_flicker_02.pdf

and

Diagnosing and Resolving LED Flicker Issues | Waveform Lighting

How do you know if there is hidden flicker in your LED bulbs? What can you do about it?

www.waveformlighting.com

 

[Unframed Dave]

Thanks.

I just found this as well. It's a bit dated as we do now have global shutter, but it still answers the question a bit more.

Colour temperature variation comes into play.

The plot thickens, well for me, anyway.

Mechanical vs Electronic Shutter: Ask David Bergman

https://www.youtube.com/user/adoramaTV Today's question: “What’s the difference between mechanical and electronic shutter on a mirrorless camera? Is mechanic...

youtu.be

 

[AlphaWorld]

Silent shutter... such a great idea until one actually uses it, indoors, in artificial light --- and it fails in the very situation it's most needed in.

It wouldn't if it wasn't for LEDs? Yes, but what lighting isn't LED these days!

Sunlight isn’t LED So there’s at least one lighting that isn’t.

More seriously:

1. The A7CR uses the same sensor as the A7RIV and V, and takes about 1/10 to scan the whole sensor. That’s the key time when running on full e-shutter.
2. The bands are due to the dimming circuit for the LEDs, not the mains frequency. The dimming circuit could be running faster or slower than mains. Still, it looks like there are 5 dark bands, and for a 1/10 second scan, that suggests the dimmer is running at 50Hz (50 bands in 1 second = 5 bands in 1/10), so it does look like it’s running at 50Hz. The dimmer has the LEDs on for a period, and off for a period (it’s called pulse width modulation) - looks like they are on for about 2/3 of the cycle and off for about 1/3, judging by the width of the light and dark bands. LEDs respond quickly, unlike incandescent lights, hence the distinct on/off bands. If the lighting were incandescent, you would not get the sharp edges to the bands.

 

[Unframed Dave]

Thank you, more learning.

Understanding that it takes a tenth of a second for the whole sensor to scan is a key factor.

Theoretical question arises then in my mind.

Why doesn't the software simply judge the exposure at the beginning of the scan and maintain it throughout? Seems like a simple solution to the banding issue.

 

[Thad E Ginathom]

Sunlight isn’t LED So there’s at least one lighting that isn’t.

I really must get out more!

 

[AlphaWorld]

Thank you, more learning.

Understanding that it takes a tenth of a second for the whole sensor to scan is a key factor.

Theoretical question arises then in my mind.

Why doesn't the software simply judge the exposure at the beginning of the scan and maintain it throughout? Seems like a simple solution to the banding issue.

It does. It’s not the exposure setting that is changing. It’s the brightness of the light that is changing.

LEDs change quickly. Old incandescent lights used to heat up a piece of wire, and it emitted light. Take the current away from the wire, and it kept emitting light for a while until it cooled down. Take the current away from an LED, and the light stops pretty much immediately. So that LED lighting is actually flickering on and off about 50 times a second. Due to the persistence of human vision, we don’t see the flicker. Persistence of vision is how come we can watch a movie at 24 frame per second and most of us don’t see it flicker.

Now as to the bands. The camera takes a shot at 1/1000, so every pixel should get light for just that long. With the chosen aperture and ISO, the top of the frame is correctly exposed (and let’s imagine that the LEDs are on when the exposure starts). All good for the first few hundred lines. But when the sensor start recording a line partway down the sensor, the LEDs turn off. The image is now under-exposed (looks dark) for the next several lines. Then the LEDs turn on again, and we get correct exposure for the next hundred lines or so. Then they turn off, underexposed, on, correctly exposed, and so on. When it takes 1/10 of a second to scan the whole sensor, a light flickering at 50Hz will give you 5 dark bands (or four, or six, depending on the exact relationship between the light flicker rate and the shutter read speed, and when the exposure started in the flicker cycle).

This can happen with a full mechanical shutter, too, but the full mechanical shutter passes over the sensor much more quickly (you know your flash sync speed? The full mech shutter takes roughly that long). So if your camera has a flash sync of 1/250 of a second, there’s a good chance that you’ll miss a dark band that only happens every 1/50 of a second. Mechanical shutters can show some banding with high frequency dimmers, like those running at 400Hz.

How is the mechanical shutter faster? The camera moves the mechanical shutter over the sensor in 1/250 of a second, and then takes as long as it needs to read the data from the sensor AFTER THE SHUTTER HAS PASSED. The shutter blocks more light from hitting the sensor, so the slow sensor can provide data on what it “saw” later. That works for both full mechanical shutter, and EFCS - both of these have the sensor blocked from light after it has had its exposure. It is only the e-shutter that is limited by the read speed of the sensor.

I hope that helps.

Edit to add: the camera cannot judge the exposure on a line by line basis - exposure is calculated for the image as a whole - if that was what you were pondering, we can dig into that.

 

[AlphaWorld]

My understanding was thats what Anti flicker (or whatever the camera manufacturer calls it) tries to stop? But I must admit the whole subject hurts my tiny brain!

A quick answer: yes, but!

A slightly slower answer:
If you have a fast enough shutter (eg: stacked sensor or mechanical shutter) then the camera can “watch” the flicker cycle, and start the exposure at the start of the next bright part. So anti-flicker can work for light flickering at, say 50Hz, and a shutter that only takes 1/250. But when your e-shutter takes 1/10 - yeah, nope!

Funnily enough there is an answer that works for a static scene (no people, no animals, no wind!) - you can use a much slower shutter speed! If you use, say, a 1 second exposure, then every pixel gets exposed to several cycles of the flicker, and it comes out as uniform illumination.

 

[Angus Cameron]

Sunlight isn’t LED So there’s at least one lighting that isn’t.

More seriously:

1. The A7CR uses the same sensor as the A7RIV and V, and takes about 1/10 to scan the whole sensor. That’s the key time when running on full e-shutter.
2. The bands are due to the dimming circuit for the LEDs, not the mains frequency. The dimming circuit could be running faster or slower than mains. Still, it looks like there are 5 dark bands, and for a 1/10 second scan, that suggests the dimmer is running at 50Hz (50 bands in 1 second = 5 bands in 1/10), so it does look like it’s running at 50Hz. The dimmer has the LEDs on for a period, and off for a period (it’s called pulse width modulation) - looks like they are on for about 2/3 of the cycle and off for about 1/3, judging by the width of the light and dark bands. LEDs respond quickly, unlike incandescent lights, hence the distinct on/off bands. If the lighting were incandescent, you would not get the sharp edges to the bands.

When you say for this sensor that the duration of a scan is 1/10, is this taken from the camera shooting at a maximum of 10fps?
If correct then the Sony A1 (30fps) is 1/30 to scan the whole sensor and the A7RM2 (5fps) is 1/5 to scan the whole sensor?

 

[AlphaWorld]

When you say for this sensor that the duration of a scan is 1/10, is this taken from the camera shooting at a maximum of 10fps?
If correct then the Sony A1 (30fps) is 1/30 to scan the whole sensor and the A7RM2 (5fps) is 1/5 to scan the whole sensor?

No.

The scan speed for the A7RIV and V has been measured, and it’s roughly 1/10. Surprising that it can do 10fps with that spec, although I believe it slows to 8fos unless you are using lossy compressed.

The A1 scans the sensor in somewhere between 1/250 and 1/270 (depending on whose measurements you trust). It is capable of shooting flash in e-shutter with a sync speed of 1/200, so there is no way it can be as slow as 1/30.

Stacked sensors have much faster sensor scan speeds. The A9 is around 1/160.

 

[Angus Cameron]

Thanks for explaining, glad I asked and didn't assume.