Pages: [1]   Go Down

Author Topic: Why I think the Sigma fpL can be a very good tripod camera  (Read 639 times)

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1914
    • http://www.guillermoluijk.com
Why I think the Sigma fpL can be a very good tripod camera
« on: April 07, 2021, 07:23:22 am »

Because it has 61Mpx, a reasonable price (much lower than A7R IV, Z7 II or R5, if purchased without the EVF) and it has ISO6, which can be used on a tripod.

When setting ISO6 the camera is likely to capture 16 shots at ISO100 (lowest genuine native ISO) in a row and average them, producing a single output RAW file. Electronic shutter should avoid any vibrations between shots keeping sharpness intact. So in essence we have a software 4-stops ND filter, but additionally averaging 16 shots means reducing noise to 1/4, or equivalently gain 2 full stops of Dynamic Range. This probably makes this camera at ISO6 be the highest DR camera in the market, making HDR bracketing (with all its drawbacks) more unnecesary than ever.

I have done a test with my Sony A7 II by averaging in the Bayer domain (I used Anton Wolf's DNG stacker for this) 16 shots taken at ISO100 over this ETTR'ed high DR scene (12 stops):
(I was surprised my camera produced some visible blooming on the chair)
http://guillermoluijk.com/misc/iso6escena.jpg

RAW histogram in stops:
http://guillermoluijk.com/misc/iso6hist.gif

ISO100 vs software ISO6 100% crop comparison (used ACR to develop the DNG files, no NR nor sharpness applied):
http://guillermoluijk.com/misc/iso6.jpg

An invisible hair on the chair in the single shots, becomes visible in the stacked image.

Want to play with the DNG files:
Regards
« Last Edit: April 07, 2021, 07:26:29 am by Guillermo Luijk »
Logged

SrMi

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 285
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #1 on: April 08, 2021, 12:02:49 pm »

I wish more manufacturers would add in-camera frame averaging. In my book, that is a more relevant feature than, e.g., pixel shift.

Sigma fp-L manual is available here: Sigma fp-L manual.

I do not understand why the manual says that with low ISO expansion "Image misalignment may occur when the subject is moving rapidly or when the shutter speed is slow." Regular frame-averaging should not be worried about movements, especially at slow shutter speeds, as the technique simulates an ND filter.

It is also weird how focus-bracketing is implemented: only 3, 5, 7, 9, 11, 13, or 15 shots are possible.
Logged

sbay

  • Full Member
  • ***
  • Offline Offline
  • Posts: 200
    • http://stephenbayphotography.com/
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #2 on: April 08, 2021, 12:09:45 pm »

I do not understand why the manual says that with low ISO expansion "Image misalignment may occur when the subject is moving rapidly or when the shutter speed is slow." Regular frame-averaging should not be worried about movements, especially at slow shutter speeds, as the technique simulates an ND filter.

I suspect it's just badly worded and they mean there may be artifacts due to time gaps between frames.

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1914
    • http://www.guillermoluijk.com
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #3 on: April 08, 2021, 12:45:23 pm »

I suspect it's just badly worded and they mean there may be artifacts due to time gaps between frames.
Agreed, in fact I'd like to check this (I'm in conversations to borrow a camera from Sigma). If the gaps are narrow enough, this is probably not an issue unless your subject includes moving lights leaving traces (vehicles at night, fireworks, lightpainting,...). So for landscape photography (clouds, leaves, water,...) this ISO6 could be the standard setting to go with this camera. Unfortunately the market doesn't seem to be paying too much attention to this model for still photography.

Regards
« Last Edit: April 08, 2021, 12:51:12 pm by Guillermo Luijk »
Logged

SrMi

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 285
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #4 on: April 08, 2021, 12:58:27 pm »

I suspect it's just badly worded and they mean there may be artifacts due to time gaps between frames.
AFAIK, the time gaps are mainly a concern with faster, not slower shutter speeds. Looking forward to Guillermo's tests and clarification :).
Logged

mcbroomf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1178
    • Mike Broomfield
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #5 on: April 08, 2021, 01:35:42 pm »

It seems like they mean object misalignment rather than the whole image (although that might also be the case if not using a tripod, depending on whether their final image aligns the frames).

Any camera can do this of course, you just then need to do the frame averaging in PS (with an action to make it easier).  Albeit PS is pretty slow.
Logged

SrMi

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 285
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #6 on: April 08, 2021, 05:13:57 pm »

It seems like they mean object misalignment rather than the whole image (although that might also be the case if not using a tripod, depending on whether their final image aligns the frames).

Any camera can do this of course, you just then need to do the frame averaging in PS (with an action to make it easier).  Albeit PS is pretty slow.
Yes, it can be done in PS, but I find that approach so cumbersome and slow that I almost never use it. On the other hand, I do it with Nikon DSLR and Olympus m43 cameras as it is done in-camera and the output is a raw file.
Logged

NikoJorj

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1065
    • http://nikojorj.free.fr/
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #7 on: April 09, 2021, 03:35:36 am »

When setting ISO6 the camera is likely to capture 16 shots at ISO100 (lowest genuine native ISO) in a row and average them, producing a single output RAW file. Electronic shutter should avoid any vibrations between shots keeping sharpness intact. So in essence we have a software 4-stops ND filter, but additionally averaging 16 shots means reducing noise to 1/4, or equivalently gain 2 full stops of Dynamic Range. This probably makes this camera at ISO6 be the highest DR camera in the market, making HDR bracketing (with all its drawbacks) more unnecesary than ever.
Thanks for your tests Guillermo!
The only downside of this stacking is that it might reveal some fixed pattern noise (that was buried under the stochastic noise before), keeping a dark frame or 16 on hand could be useful.

Just a quibble : aren't the constraints of this 16-image capture quite similar to those of bracketing (your other beyond-zero-noise example shows excellent results too)? I hate tripods...  :-[
Logged
Nicolas from Grenoble
A small gallery

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1914
    • http://www.guillermoluijk.com
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #8 on: April 09, 2021, 05:46:55 am »

Thanks for your tests Guillermo!
The only downside of this stacking is that it might reveal some fixed pattern noise (that was buried under the stochastic noise before), keeping a dark frame or 16 on hand could be useful.

Just a quibble : aren't the constraints of this 16-image capture quite similar to those of bracketing (your other beyond-zero-noise example shows excellent results too)? I hate tripods...  :-[
Sure, if pattern noise becomes visible a good idea is to substract a positive dark frame or average of several darkframes, maybe I'll do that later to try to map my sensor's pattern noise. If you look carefully into my example, there is a very thin hair over the chair just behind Mazinger Z that became visible only after stacking. Cleaning images can reveal undesired things that remained hidden.

The constraints of bracketing for HDR are much higher than in average stacking. Think that if you are just averaging some RAW files, any camera movement will translate into motion blur, but this is no different to shooting at slow speeds with an unstable camera, and the resulting RAW file will decode finely (blurred). On the opposite, any missalignent for HDR, if done in the Bayer domain can lead to terrible demosaicing artifacts. That's why I find impractical to create HDR RAW files as a regular workflow, mine was just an experiment and I have to say I was surprised at how well the shots remained aligned at pixel level. Probably because of this no camera make has risked to build in-camera RAW files so far.

Regards

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1914
    • http://www.guillermoluijk.com
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #9 on: April 09, 2021, 08:21:27 am »

I shot 16 darkframes @ ISO6400, and averaged them to show spatial noise patterns. The data on the 16 RAW files was first averaged at their original size. Then I rescaled to 763 pixels width to make patterns more visible over shot noise, and it was then when I substracted the bias level (512). To the linear normalized B&W version, a colour (magma) version was added to identify more easily sensor noise areas. All the process was done using Sony's ARW files since the DNG conversion done by Adobe's DNG Converter automatically substracts the bias level on each RAW file, which is a big disadvantage to get rid of some random noise.

http://guillermoluijk.com/misc/darkframe.png

It seems my sensor has a lot more noise in the lower part (upper part in the camera), and a specially low noise area on the right (left side in the physical camera).
There are both horizontal and vertical pattern noise, by vertical is by far more important.

Regards

sbay

  • Full Member
  • ***
  • Offline Offline
  • Posts: 200
    • http://stephenbayphotography.com/
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #10 on: April 09, 2021, 11:52:27 am »

So on Kasson's blog he measured the readout time of the a7R4 electronic shutter at about 1/10s. I think that's long enough to cause artifacts with moving objects even in landscapes -- for me this would be something like a tree branch moving in the wind, a wave coming in during a seascape, or a boat slowly rocking with the waves. Not entirely sure though, I suspect it depends on whether the object would move enough in the 1/10s so that the ghosts did not overlap.

If it could be implemented so that most artifacts were avoidable, it would really be a game changer for a lot of landscape photography. You have up to 4-stop ND built into the camera (compared with iso 100) and way less noise -- sometimes I still exposure blend but doubt I would bother with ISO 6.

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1914
    • http://www.guillermoluijk.com
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #11 on: April 09, 2021, 12:47:30 pm »

So on Kasson's blog he measured the readout time of the a7R4 electronic shutter at about 1/10s. I think that's long enough to cause artifacts with moving objects even in landscapes -- for me this would be something like a tree branch moving in the wind, a wave coming in during a seascape, or a boat slowly rocking with the waves. Not entirely sure though, I suspect it depends on whether the object would move enough in the 1/10s so that the ghosts did not overlap.

If it could be implemented so that most artifacts were avoidable, it would really be a game changer for a lot of landscape photography. You have up to 4-stop ND built into the camera (compared with iso 100) and way less noise -- sometimes I still exposure blend but doubt I would bother with ISO 6.
In DPreview I read fp's readout it 1/10s as well. But still I'm not sure why this should be an issue for landscapes. Don't think of 1/10s as the exposure time, it is just the time needed to scan the whole sensor. A leaf in a tree will be read in a very small fraction of this 1/10s so you're not likely to have any strange deformations nor ghosting.

In this rolling shutter simulation (sensor readout 1/60) a quick movement like this tennis racket gets hardly deformated. Tree leafs are much smaller and move more slowly so 1/10 should suffice:

http://guillermoluijk.com/tutorial/rollingshutter/raqueta.gif

I have a feeling this ISO6 will just struggle with moving lights, but is fine with any other moving natural subject.

Regards

sbay

  • Full Member
  • ***
  • Offline Offline
  • Posts: 200
    • http://stephenbayphotography.com/
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #12 on: April 09, 2021, 02:30:42 pm »

So I'm not thinking of deformation within a single shot caused by a slow readout on the electronic shutter but rather the gap between successive shots will cause moving objects to appear distinctly multiple times instead of having a smooth blur.

For example, I've attached a shot done with smooth reflections (in camera averaging app on a sony a7R2) taken at a fairground. If you look at the crop, you can see a walking person leaves distinct "ghosts" for every frame averaged in the stack. With a normal ND filter and a single long exposure, the person would leave a smooth blur trail.

Now smooth reflections only works with a regular shutter, not electronic, so maybe the gap between shots would be smaller with e-shutter. But I suspect there's always going to be some combination of movement speed and shutter speed that will leave multiple impressions instead of smooth blur.

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1914
    • http://www.guillermoluijk.com
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #13 on: April 09, 2021, 04:35:39 pm »

Moving lights, that is exactly what I was saying. But there are no moving lights in landscape, arquitecture or indoor pictures. That was my point.

https://www.sigma-imaging-uk.com/lounge/paul-monaghan-122019/

"The ISO 6 mode is particularly worthy of attention as it allows you to lower the amount of light in the scene and works really well particularly with landscapes. Itís like having a free 5 stop ND filter at hand but does require the use of a tripod. To test this, I did a comparison between ISO 6 and an ND filter at the same shutter speeds."

One is ISO200 with 6 stop Benro ND Filter and the other is ISO6 no ND filter.


Unfortunately this scene would have been great to test at ISO6 to check for gaps, but he didn't:


Regards

sbay

  • Full Member
  • ***
  • Offline Offline
  • Posts: 200
    • http://stephenbayphotography.com/
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #14 on: April 11, 2021, 06:02:00 pm »

Although the person is lit from the side by the tent lights, he is not that much brighter than his surroundings. So I don't think is the same as a moving light such as a car trail, firework, or carnival ride (which is I what I thought you meant).

But here's basically the same example in all natural case. It's simply a brighter subject moving against a darker background. If you look at the ends of the reeds, you can see the individual frames being averaged.

The reverse can also happen too -- a dark object moving in front of a brighter background can leave traces.

sbay

  • Full Member
  • ***
  • Offline Offline
  • Posts: 200
    • http://stephenbayphotography.com/
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #15 on: April 11, 2021, 06:06:14 pm »

This example is from shooting out of a sea cave. If any situation would warrant ISO 6, this is it due to the extreme brightness difference.

Here there are two types of artifacts. The first is the odd dark shadow edge of the rock which is repeated multiple times.

The second are the bubbles floating through the water which get repeated whereas a single shot would result in a smooth blur. Based on the number of bubbles, I must have done an 8 stack in smooth reflection (unfortunately sony doesn't record the number of frames averaged in metadata). The line across the crop I think is from vignetting correction (I think it's baked into the smooth reflection app).

In this case, I think these artifacts would be easy to fix. But still, it wouldn't be necessary to check or fix with an ND filter as opposed to an averaging scheme. The reeds I think would be harder to correct -- not really sure how I would do that as simply removing by cloning them out wouldn't look right.

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1914
    • http://www.guillermoluijk.com
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #16 on: April 12, 2021, 07:13:54 pm »

I understand your point. On all those examples exposure time is longer than the gap between shots. I expect the case with Sigma's ISO6 is the opposite; if the gap remains much lower than the expposure times, that kind of movement will be hardly visible (with moving lights being perhaps the only the exception).

It's all about achieving a high exposure/gap ratio:

- = gap
O = exposure

Worst case scenario:
---------------OO---------------OO---------------OO---------------OO---------------OO

Preferred situation:
OOOOOOOOOOOOOOO--OOOOOOOOOOOOOOO--OOOOOOOOOOOOOOO--OOOOOOOOOOOOOOO

In these 16 averaged captures (software 4-stops ND filter) there were no issues with moving water:

http://www.guillermoluijk.com/misc/ndsimple.jpg

http://www.guillermoluijk.com/misc/nd4pasos.jpg

http://www.guillermoluijk.com/misc/ndsoftwarecomp.jpg


One more example: 5-stops software ND filter

http://www.guillermoluijk.com/misc/nd_original.jpg

http://www.guillermoluijk.com/misc/nd_final.jpg

Regards

« Last Edit: April 12, 2021, 07:19:05 pm by Guillermo Luijk »
Logged

sbay

  • Full Member
  • ***
  • Offline Offline
  • Posts: 200
    • http://stephenbayphotography.com/
Re: Why I think the Sigma fpL can be a very good tripod camera
« Reply #17 on: April 13, 2021, 11:24:32 am »

Yeah I definitely hope the sigma can get the gap down. On my sony a7r2, I think the time gap was at least 200ms which is very long time. Looking forward to seeing your results when you get a chance to test the sigma.

When I've used averaging around water, I generally have the most success when the flow is regular like a stream or waterfall or if I'm trying to flatten water completely. But it would be really great if it would work at the ocean -- here I typically try for a total exposure time around 0.5s which keeps the rough shape of a wave but adds a bit of motion blur. The water doesn't exactly repeat so getting a smaller time gap would be more critical.
Pages: [1]   Go Up