Horizontal lines (Shutter Speed incompatibility?)

Prokofiev · Feb 18, 2023

Hello. I recorded some videos on a rock concert and we can see horizontal lines, especially in bright zones with a certain contrast. Here is a sample:

Here is a frame with clear lines:

I shot this with a Fujifilm X-T3 at 25 fps and 1/50 shutter speed. With in-camera B&W simulation (but I recorded other takes in color and it is the same, although in black and white is more noticeable).

Is it about the shutter speed? Is the SS incompatible with the lights (one light or some of them)? Should these lines disappear if I use 1/30 or 1/60 maybe?

I tried a plugin called Flicker Free. But it is too slow and I don’t have a separate GPU, just the one from the Intel CPU, so it is hard to try different options quickly and it takes a lot of time to render. When I try, it doesn’t seem to do much in this case. Maybe it reduces the contrast of the lines, I don’t know.

Do you know a way to eliminate or reduce these lines?

Andrew S10 · Feb 18, 2023

Were you in a region with 50Hz electrical frequency refresh rate?

The brightness of the lights might be controlled through pulse-width modulation, which can be challenging to work around.

Does your camera's menu have any Flicker Free or Synchro Scan mode?

Try duplicating the clip, stacking one on top of the other with 1 frame offset, setting the opacity to 50%, and changing the blend mode to Add. If that doesn't work, you might need to add a third or fourth layer, dividing the opacity by the number of clips. This method will introduce ghosting, so choose your poison.

The de-flicker pulg-in will probably run faster if you first transcode the H.265 clip to ProRes, DNxHR, or CineForm.

Prokofiev · Feb 19, 2023

Andrew S10 said:
Were you in a region with 50Hz electrical frequency refresh rate?

Yes, I think so.

When these lines show in videos, does it mean an incompatibility of SS and frequency refresh rate? Or is it about an specific characteristic of these stage lights (and probably the lights going directly to the lens)?

Andrew S10 said:
The brightness of the lights might be controlled through pulse-width modulation, which can be challenging to work around.

How is that? Do you mean controlling the lights? (in this case, that was not possible)

Andrew S10 said:
Does your camera's menu have any Flicker Free or Synchro Scan mode?

I don't think so.

Andrew S10 said:
Try duplicating the clip, stacking one on top of the other with 1 frame offset, setting the opacity to 50%, and changing the blend mode to Add. If that doesn't work, you might need to add a third or fourth layer, dividing the opacity by the number of clips. This method will introduce ghosting, so choose your poison.

Thank you for the recommendation. I tried it and it doesn't eliminate the lines, if that was the purpose.

Also, I can't find the Add mode in Premiere pro. Overlay, "Linear dodge (Add)" ... but not Add.

Anyway, I don't see how this method would eliminate the lines.

Andrew S10 said:
The de-flicker pulg-in will probably run faster if you first transcode the H.265 clip to ProRes, DNxHR, or CineForm.

OK. That is a good idea, thank you.

John Vickers · Feb 21, 2023

Prokofiev said:
Hello. I recorded some videos on a rock concert and we can see horizontal lines, especially in bright zones with a certain contrast. Here is a sample:

Here is a frame with clear lines:

I shot this with a Fujifilm X-T3 at 25 fps and 1/50 shutter speed. With in-camera B&W simulation (but I recorded other takes in color and it is the same, although in black and white is more noticeable).

From CineD, the X-T3 has about 9ms rolling shutter at 4kP25.

There are about 20 bands in the still image, so the lighting seems to be strobing at about

(20 cycles) / (9ms) = 2200Hz

About 44 lighting cycles per exposure time.

I write "strobing", because if the light just varied a bit - say from 60% to 100% over a cycle, that would get averaged out by the exposure time. Also, if the lighting was at 100% for 90% of a cycle, and 0% for 10% of a cycle, that would get averaged out too.

I think what's happening is that most lines are exposed with 44 strobe pulses, and a few lines - the darker lines - with 43 strobe pulses (those numbers might not be exactly right, but hopefully the idea is clear).

The other weird thing about the banding is that it fades in and out with a period of about 2 seconds. I *guess* that is caused by frequency drift in the lighting. The frame rate and exposure time of the X-T3 should be stable to about 20ppm or better.

Prokofiev said:
Is it about the shutter speed? Is the SS incompatible with the lights (one light or some of them)?

If what I've written is on the right track, it's more that the light(s) causing the problem are incompatible with any shutter speed.

Shooting at 1/25 s shutter speed should attenuate the dark lines by brute force - it should halve the difference between the light and dark lines, compared to 1/50s, by averaging over twices as many lighting cycles.

Prokofiev said:
Should these lines disappear if I use 1/30 or 1/60 maybe?

Don't think that will help.

Prokofiev said:
I tried a plugin called Flicker Free. But it is too slow and I don’t have a separate GPU, just the one from the Intel CPU, so it is hard to try different options quickly and it takes a lot of time to render. When I try, it doesn’t seem to do much in this case. Maybe it reduces the contrast of the lines, I don’t know.

Do you know a way to eliminate or reduce these lines?

I suspect the answer is "change the lighting". If you have access to the location, maybe you could take your X-T3 and an HDMI monitor (and/or a laptop), and experiment with the lighting.

2kHz strobing may be visible to the naked eye: try waving a drumstick under this lighting. Did the drummer notice any strobing?

Entropy512 · Feb 21, 2023

Prokofiev said:
Andrew S10 said:

Were you in a region with 50Hz electrical frequency refresh rate?

Click to expand...

Yes, I think so.

When these lines show in videos, does it mean an incompatibility of SS and frequency refresh rate? Or is it about an specific characteristic of these stage lights (and probably the lights going directly to the lens)?

It is more an issue of rolling shutter vs. light pulsing. Sometimes matching shutter speed to light frequency exactly can reduce/mitigate these artifacts, but not always.

Prokofiev said:
Prokofiev said:

The brightness of the lights might be controlled through pulse-width modulation, which can be challenging to work around.

Click to expand...

How is that? Do you mean controlling the lights? (in this case, that was not possible)

He means the method by which the lights themselves dim - unfortunately, there's not much you can do about lights that behave like this unless the venue owner is willing to change lighting to support a video shoot. One thing to keep in mind is that if this was a paid shoot, you should inform clients that results cannot be guaranteed if you don't have control over lighting.

https://en.wikipedia.org/wiki/Pulse-width_modulation - most LEDs are controlled in this manner, because reducing current significantly changes the efficiency of the LED and thus can make it difficult to dim predictably, or in the case of white LEDs, can shift the color of the LED.

Unfortunately it is likely with modern LED stage lighting that the lighting is no longer pulsing at line frequency, which makes things worse for any camera operators. As an example, it was estimated from the rolling shutter artifacts generated by a Philips Hue RGB LED bulb and known properties of the camera in question that they PWM at around 600 Hz, not 50 or 60. (Sorry, can't find the link to that post now, as it's years old and DPR's search function is pretty garbage.)

John Vickers · Feb 23, 2023

Andrew S10 said:
Were you in a region with 50Hz electrical frequency refresh rate?

The brightness of the lights might be controlled through pulse-width modulation, which can be challenging to work around.

Does your camera's menu have any Flicker Free or Synchro Scan mode?

Try duplicating the clip, stacking one on top of the other with 1 frame offset, setting the opacity to 50%, and changing the blend mode to Add.

If that doesn't work, you might need to add a third or fourth layer, dividing the opacity by the number of clips.

A (usually) better family of simple filters that operate over more than two samples are the binomial filters, which have relative weights from Pascal's triangle , and approach gaussian filters as the number of samples gets large.

For two samples (in this case frames), we have weights: 1/2, 1/2

For three, we have weights: 1/4, 2/4, 1/4

For four: 1/8, 3/8, 3/8, 1/8

For five: 1/16, 4/16, 6/16, 4/16, 1/16

...

These schemes have been quite popular in hardware and DSP, when speed/simplicity/parallelism are at a premium.

Andrew S10 said:
This method will introduce ghosting, so choose your poison.

If one knew in advance that this crazy frame averaging was going to be necessary, then shooting at a shutter speed equal to (or very close to) the frame rate - i.e. 360° shutter would minimise the ghosting, by blurring the individual frames more.

Then that could cause other flicker problems, say if the shutter speed had been chosen to eliminate mains-related flicker.

In this particular case, the OP would have more flexibility in post-processing if they shot 50P, 1/50 s, rather than 25P, 1/50 s.

If the OP had wanted to create 24P somewhere with 60Hz mains electricity, out of an abundance of caution, they might want to shoot at 48P, 1/60 s (288°), rather than 24P, 1/60 s (144°).

Then if there aren't any weird problems requiring filtering across frames, just drop every other frame, giving a conventional sense of motion at the target frame rate. (I think this has been mentioned on this forum before).

But there aren't a lot of low-budget cameras that support 48P (or 47.952P), particularly with sound, though that seems to be changing.

Andrew S10 said:
The de-flicker plug-in will probably run faster if you first transcode the H.265 clip to ProRes, DNxHR, or CineForm.

Djehuty · Feb 25, 2023

Cheap LED lights don't follow the 50/60hz PAL and NTSC standards, their refresh rates are all over the place because that's how they control brightness.

Sean Nelson · Feb 25, 2023

Djehuty said:
Cheap LED lights don't follow the 50/60hz PAL and NTSC standards, their refresh rates are all over the place because that's how they control brightness.

My car has an electronic rear view mirror that provides a much wider field of view than a conventional optical mirror does. I'm amazed at the variety of different strobe rates I see on the LED headlights of the cars behind me. The first time or two I thought a cop was trying to pull me over...

Prokofiev · Mar 18, 2023

Hello. Thank you for the comments.

It is not easy to follow the math if I don’t know the principles. So, every rolling shutter lasts 9ms at 25 fps (and at 1/50 shutter speed? Or at ANY shutter speed?).

A shutter speed of 1/50 lasts 20 ms (milliseconds, right?).

So… does every shutter cycle have like two 9ms sub-cycles?

You don’t have to explain me in detail; I just want to know if I am thinking right about this.

And why are these cycles represented as equidistant lines across the image…?

Anyway, I recorded one of the takes accidentally at 1/125 shutter speed. In this take, there is another kind-of line/bands, more like rolling/vibrating bands. Plugins like Flicker Free and Deflicker can eliminate magically these bands, especially when there isn’t two much movement. When there is movement, there are options to take charge of the problem. But if there is too much movement, like in rock concerts with vibrating lights, handheld camera, and musicians moving rapidly, artifacts appear.

But these plugins don’t eliminate nor reduce the black and steady lines I mentioned in my first post, recorded at 1/50 SS. Or at least I haven’t found an option yet.

Here is an example of effective deflicker plug-in. In this case, I shot at 60 fps and 1/60 SS. There were tube lights, and as I said I live in a 50hz country (BTW, I shot also at 60fps with some LED regular bulbs and there was no flicker, but old tubes were always problematic, right?) .

And in this case, the one I mentioned before, I recorded at 1/125 SS. You can see the rolling bands on the left side, and the black and thin lines on the right. The plug-in eliminates the rolling bands, but not the black and thin lines.

Entropy512 · Mar 20, 2023

Prokofiev said:
Hello. Thank you for the comments.

It is not easy to follow the math if I don’t know the principles. So, every rolling shutter lasts 9ms at 25 fps (and at 1/50 shutter speed? Or at ANY shutter speed?).

A shutter speed of 1/50 lasts 20 ms (milliseconds, right?).

So… does every shutter cycle have like two 9ms sub-cycles?

You don’t have to explain me in detail; I just want to know if I am thinking right about this.

And why are these cycles represented as equidistant lines across the image…?

Anyway, I recorded one of the takes accidentally at 1/125 shutter speed. In this take, there is another kind-of line/bands, more like rolling/vibrating bands. Plugins like Flicker Free and Deflicker can eliminate magically these bands, especially when there isn’t two much movement. When there is movement, there are options to take charge of the problem. But if there is too much movement, like in rock concerts with vibrating lights, handheld camera, and musicians moving rapidly, artifacts appear.

But these plugins don’t eliminate nor reduce the black and steady lines I mentioned in my first post, recorded at 1/50 SS. Or at least I haven’t found an option yet.

Here is an example of effective deflicker plug-in. In this case, I shot at 60 fps and 1/60 SS. There were tube lights, and as I said I live in a 50hz country (BTW, I shot also at 60fps with some LED regular bulbs and there was no flicker, but old tubes were always problematic, right?) .

And in this case, the one I mentioned before, I recorded at 1/125 SS. You can see the rolling bands on the left side, and the black and thin lines on the right. The plug-in eliminates the rolling bands, but not the black and thin lines.

Rolling shutter and shutter speed are two different things. Above a certain shutter speed, the leading and trialing shutter curtains (whether mechanical or electric) travel across the sensor as a slit. As the effective shutter speed gets faster, the slit gets narrower.

A related phenomenon that technically applies to electronic shutter, but in practicality rarely does, is the flash sync speed - for either a mechanical shutter or electronic, this is the speed at which there is a period in which the entire frame is active at the same time. It's typically MUCH lower (frequently unusably low) for non-stacked electronic shutters.

That flash sync speed is related to how non-constant illumination sources behave - because they're similar in behavior to a flash going off over and over and over again faster than the human eye can percieve, but such that a camera CAN see.

basyl1891 · Mar 21, 2023

Prokofiev said:
Plugins like Flicker Free and Deflicker can eliminate magically these bands, especially when there isn’t two much movement. When there is movement, there are options to take charge of the problem. But if there is too much movement, like in rock concerts with vibrating lights, handheld camera, and musicians moving rapidly, artifacts appear.

No magic here, it's just some sort of frames averaging. Naturally, movement makes it problematic. There are some ways to compensate it, but only to a certain degree.

Prokofiev said:
And in this case, the one I mentioned before, I recorded at 1/125 SS. You can see the rolling bands on the left side, and the black and thin lines on the right. The plug-in eliminates the rolling bands, but not the black and thin lines.

There are at least 2 sources of light, with different frequencies. Again reminder what deflicker does - it averages frames in a way. So if flicker is fast it's easy to fix, bc you only have to avg 2-3-4 frames. But if the bands barely moving, like on the right, to average that you need dozens-to-hundreds of frames. And that is not feasible for non-static scene

Search

Horizontal lines (Shutter Speed incompatibility?)

Prokofiev

Well-known member

Andrew S10

Senior Member

Prokofiev

Well-known member

John Vickers

Senior Member

Entropy512

Veteran Member

John Vickers

Senior Member

Djehuty

Senior Member

Sean Nelson

Forum Pro

Prokofiev

Well-known member

Entropy512

Veteran Member

basyl1891

Well-known member

Similar threads

Keyboard shortcuts