FS7: Chromatic errors at sparks

A

A7s_user

Active member
Sorry for this noob question again, again with my topic "fireworks displays with FS7" - but who could guide me better, than you professionals, please?

I am still searching for the correct settings, to get good bright (and the correct) colors as well as fine lowlight charcoal gold.
Few days ago I've filmed an very small display just with consumer fireworks (one is thankful for even those tiny displays to have another trial and error) in Slog2 in Custom mode.

Detail was off, as well as Skin Detail, but Aperture was on with +-0.

Please check the 3 frames.
01: bright silver comets with titanium sparks
https://dl.dropboxusercontent.com/u/49433518/01.jpg

02: popping flowers/dragon eggs with silvery / bright golden sparks
https://dl.dropboxusercontent.com/u/49433518/02.jpg

03: some golden glitter spiders with a silver titanium whirl below
https://dl.dropboxusercontent.com/u/49433518/03.jpg

Images are graded with LUT, but the problem is obvious without grading.

All effects have sparks. You can see, as darker they become (the top of the comet is very hot and bright, the sparks will darken and cool down immediatly in its tail, golden is always not bright (spiders and popping flowers), the whirl was at the end of its flight) the more they don't have their correct color (should be pure silver or gold), but are extremly colorful, lots of tiny blue, red, orange and violet sparks. They should be absolutly not like that.

But: as soon as the image is slightly out of focus (not in the screens but seconds later, as I changed focus at same fov), the colored sparks are gone and they have their correct silver/gold. So it seem to be related to detail/focus.
I filmed with brand new Tokina 11-20mm on Novoflex Nikon adapter, at 11mm. Later I've zoomed a little, the effect was gone - but I don't know becaue auf fov or because I was out of focus.

What could it be, please? Some detail thing within the camera or just the lense? As I have an important display in two days, I have no chance to test before again...

Any hint is highly appreciated, thanks in advance.
 
Last edited:
Well, beyond lens issues, you might look at your codec and such. Look into how color space works. 4:2:2 color space gives each pixel location a unique luma value, but shares color values across two pixels, and your fireworks show the most color issues when they're just one or two pixels wide. So it looks like it may be grabbing the color value of the night sky, then applying the brightness value of fireworks. That sort of thing.

Shooting 4k and downsampling to 1080 4:4:4 carefully might help.

I'd really choose the very best codec you can for this task.
 
Fireworks are way beyond the dynamic range of the camera even in log mode. Defocusing lessens the exposure from over to somewhere within the cameras dynamic range bringing the color back. In my past motion picture work we would test and find an exposure that would be the most pleasing, but never got close to perfect.
 
Thank you both for your comments!

@Viddovation
This was shot with FS7 in XAVC-I UHD/25p, so the best I can internally get out of the camera (besides 50p). I do film fireworks only now for almost 15 years and never had such an intense experience. I do film with an A7s too, sometimes in HD internal recorded and sometimes in UHD/25p via HDMI on Odyssey7Q+ recorded. There is not such an effect, a least far, far, far less. I assume, this is a "special feature", at least with my settings within the camera. The FS7 does show too very hard color edges at the normal colored stars - which look much to sharp/anormal, compered to my A7s and EX1R:

https://dl.dropboxusercontent.com/u/49433518/04.jpg
Disclaimer: This is not the kind of displays I normally film ;), it's just a test

@banolan
But the very bright top parts of the white comets (image 01) are perfect, its always the cooler/darker sparks which show the problems.
 
A7s_user said:
(...) All effects have sparks. You can see, as darker they become (the top of the comet is very hot and bright, the sparks will darken and cool down immediatly in its tail, golden is always not bright (spiders and popping flowers), the whirl was at the end of its flight) the more they don't have their correct color (should be pure silver or gold), but are extremly colorful, lots of tiny blue, red, orange and violet sparks. They should be absolutly not like that.

But: as soon as the image is slightly out of focus (not in the screens but seconds later, as I changed focus at same fov), the colored sparks are gone and they have their correct silver/gold. So it seem to be related to detail/focus.
I filmed with brand new Tokina 11-20mm on Novoflex Nikon adapter, at 11mm. Later I've zoomed a little, the effect was gone - but I don't know becaue auf fov or because I was out of focus.
Click to expand...


What you are seeing is a common digital imagining artifact, color aliasing. It's been tested that FS7's low pass filter is optimized for achieving higher sharpness at an expense of slightly greater likelihood of moire / aliasing. It is not a problem in most situations, especially in 4k and UHD resolutions where the OLPF is the most effective. In your case the high brightness sparks in combination with a high resolution lens cause the sensor to struggle. Lens rendered details hit the threshold of OLPF's effectiveness (individual spark are close to or at resolution limit of the sensor and cause artifacts). When the lens is slightly defocused it no longer happens because sparks are rendered as bigger disks which don't excite aliases. Same with zooming in, the individual sparks became bigger (you can think about resolving power of the lens and resolving power of the sensor as frequencies - as you changed the frequency of the rendered details by zooming in, it no longer caused interference with the frequency of the sensor). You may want to try using a slightly diffusing filters. I'd suggest Tiffen Digital Diffusion 0.25, 0.5 or 1 filters but in your case they may introduce double images, ghosts of bright fireworks. Netting the rear of the lens might be a solution. This page shows extreme examples of the same phenomenon, it is rendered by a very old digital camera but the principles are the same. Look what happens with specular highlights of the water in the picture of the ship (scroll down): http://www.robgalbraith.com/multi_page08dd.html?cid=7-6463-7191-7457
 
Last edited:
Thank you, @broughtonfilm. Your explanations sound so logical and understandable. And they came just right before a really large display today. So, better not to try to find the perfect focus, otherwise all the titanium and aluminium sparks will be rendered with massive color aliasing. Whow, that would be one more reason not to keep the FS7 for my special needs. Thanks a lot, this really helps, just in time!
 
As one who's shot quite a bit of fireworks, my impression is that your examples are 2-3 f-stops overexposed. The 'dark sky' and the smallness of the points of light makes it appear too dark in the viewfinder. The usual reaction is to want to open the aperture. But then you find out, when looking at it on a full resolution monitor that it's overexposed. I've noticed that with most any modern camera, overexposure exacerbates all of these aliasing problems. Fireworks look best when you protect the higlights.

https://youtu.be/W5BdsoyhA7A?t=6m48s

(Shot on my EX1 back in 2010)
 
@basspig
Thanks again for your input. The sample screens above were taken almost at daylight. That's why the sky is that blue (before the rough grading it was even much brighter). Remember, the color aliasing happened with lowlight gold too, so overexposure is not the main guilty, but probably the sharpness, as @broughtonfilm suggested. But its true, its very difficult to catch the different bright colors as well as the lowlight charcoal gold. And I don't wont to kill all the details from the scenery and sky. That's why I do regulate aperture manually all the times during the shows since 15 years.

Meanwhile I tested more and improved my grading and become slowly somehow more happy with the results out of the FS7. But I still beg for a FS7 like cam with Sony 3CMOS in 4k, that would be a rather easy play. As it was with the EX1 and EX1R.
 
Yeah, I'm rather hesitant to buy a single imager camera myself, but it seems there is no other choice!

Definitely avoid sharpening in-camera when shooting. Any electronic sharpening ruins this kind of image.

Who shoots fireworks in daylight? I've never seen a show begin before 9:40PM out here in New England. The western sky is only a little bit lighter than pitch black, but it's full night when the show starts.

If your sky is appearing that bright (unless it was 7PM in May) then you're letting in too much light.

I shot the fireworks show with ND1 at F11, to give you an idea of the settings needed to preserve the highlights. Each spark is as bright as the sun, just tiny in area, so we perceive it as much less intense, so basically, you would use the same settings as you would use for shooting a solar eclipse.
 
It was an opening of a supermarket, so just some tiny bangs for the customers at 9PM at the beginning of May. Nothing I normally would drive for, but I need to test my FS7 more ;)
As mentioned in another thread I don't want to kill the ambience with ND + F11. In my opinion this produces dead fireworks, you loose all charcoal sparks, the reflections in the smoke and all details of the scenery, the sky and the audience. But this a topic we better discuss privatly, it won't be interesting for the others.

There will be a 1/2'' 3CMOS 4K sensor by Sony, I am pretty sure. They developed the 2/3'' 3CMOS 4K already, so there is a need and they do further develop 3CMOS, so there will be (cross fingers) in near future. Question is "when" and will it be a cam with eMount to have the same lense possibilities? I would highly vote for a PXW-X300 with the same codecs + XAVC-I with 3CMOS 4K, SxS/XDQ + eMount.


Back to topic:
What I don't understand is: why does one have to protect the very highlights in the single sensors FS7 + A7s so hardly, with using slog gammas with 12-14 stop dynamic range and still having massive problems with clipping & shifting of extreme highlights - while rather old cams with 3CMOS sensors and gammas with only 5-6 stops range easily produce awesome fireworks colors, even if you would keep the lense fully open. Of course the image would be totally blown out and not usable, but you still would have awesome colors. And with F2.4-4.8 on an EX1R you have a perfect image, great colors - and no hurting/really noticable highlight clipping, like it happens with the modern single sensor cams. I simply don't get the point.
 
highlight aliasing, this effect can be seen on all compressed recordings (F5/FS7), it's caused by the way the camera debayers the signal.

Only way to avoid this is to shoot RAW.
 
Does it look like the fireworks sample I posted above looks "dead" with no ambiance? That was shot F11 with ND1. I think perhaps you're asking too much of current camera technology. You want even more dynamic range than the human eye can handle at once. But then there's no display in existence that can reproduce that anyway. In fact, in my experience, display technology is a limiting factor here, more so than the cameras.

With projection, you'd want 25K lumens for a 10' screen to get that kind of intensity, but with commercial projectors having only 2000:1 contrast ratio, you would not have black--you would have grey illuminating the room where black is supposed to be displayed. But the whites would be at the proper intensity.

The older cameras probably had poor optics and blurred the image over larger circles, and that caused the operator to stop down a bit. But I haven't seen any good fireworks videos shot on older cameras--they're all blown 6-7 F-stops over and look washed out.

I think the EX1 looked just as bad when it was over-exposed. True, there's an advantage with 3 imagers, in that color separation is absolute and aliasing is irrelevant in such a setup, but it should be possible to do this right with today's single sensor cameras.
 
You must log in or register to reply here.
Back
Top