Thread 4437489

What's the excuse for literally all mirrorless bodies not having built in ND filters as a feature?

>>4437489 (OP)
I want a built in circular polarizer with a dial on the back that spins it around, DAMMIT.

>>4437489 (OP)
I think all mirrorless should have graduated ND filters that let you adjust both angle height of the graduation. Why has NO ONE DONE THIS FOR ME YET?

because it adds one to two pounds of weight and a great deal of manufacturing complexity

>>4437512
>Every camera manufacturer when Sony first had IBIS: our cameras dont have IBIS because it adds one to two pounds of weight and a great deal of manufacturing complexity
>now every camera body has IBIS

>>4437489 (OP)
I'm simply against computational photography

>>4437529
it also adds a fuck ton of volume no matter what
look at every single camera with an internal ND filter

Japan has crazy patents. OM has probably patented this as a way of emulating NDs. The limitation on OM is that the shutter must be rather slow. You cannot do live ND at say a shutter speed of 800.
Sigma has this on the FP cameras, but better. Instead of being for emulating NDs, it is simply a lower ISO. This means dramatically good shadow performance. The problem of course is that the FPs have no IBIS, so you have to bring along a tripod.
But on OM with MFT sensors the advantage of low ISO emulation for noise would be massive. I've tested this inside where shutter was quite slow and it worked really well, virtually noiseless images. You can of course emulate Sigma-style low ISO on any camera with burst mode and averaging frames (termed AHDR), or by doing conventional HDR (which your camera probably has a mode for anyway). You could even use frame averaging to emulate an ND, but it'd be a huge pain and I'd rather just carry a filter or two.

>>4437547
It's computational photography
>As it happens, the actuality of how Live ND works is much simpler. The sensors used are capable of 60 frames-per-second readout, so the Live ND, which only operates for shutter speeds slower than 1/30th second, simply takes multiple short exposures and composes them in the image processor. The amount of ND depends on the relationship between the length of the short exposures and 1/60th second. So, for instance, if 1/120th second exposures are used, it will produce the effect of a 50% ND filter.

It's been around for a while.

>>4437547
ricoh gr has it and it is pretty smol
>>4437555
at least on the foveon sigmas iso 50 is just iso 100 with lower gain so you lose 1 stop of highlights, i.e. you might as well just underexpose 1 stop and compensate in post
dunno how it is on the fp tho tbqh

>>4437546
soon snoy with have DALL-E AI and midjourney in their cameras. You will take a picture of a scenario and with voice command prompts say how you'd like the focused object to perform, howd you like the background, lighting, if fantasy if cartoony. CLICK
The camera will make and take an AI representation of the scene and your prompts.
Have fun!

>>4437498
You could get the EF-RF drop-in filter adapter.

>>4437773
You'd be stuck with inferior DSLR lenses.

>>4437498
but then forget weather sealing... unless you make that dial communicate wirelessly instead of machanically.. hmm capacitive or inductive sensors would be cool. Giving the user somethign haptical/mechanical feeling but the communication is wireless based. I just hope no touch sensors

>>4437555
This is probably the real answer.
>>4437547
This is about live ND, aka, e-shutter internal exposure blending. Not physical internal mechanized ND filters that pop up in front of the physical sensor.

>>4437498
I prefer a linear polarizer

>>4437546
All digital photography is computational photography you fucking retard.

>>4446465
demosaicing a bayer array and then hitting it with denoise and unsharp mask is a lot less fake than stacking a gorillion exposure while applying AI inpainting that can recognize and replace objects

>>4446311
What is the difference?

>>4446468
No, it isn't.
You are wrong and don't even understand why you're wrong.
Photo stacking is essentially "purist" grade. There's really nothing "computational" about it.

Imagine for a moment the top left most photosite of your sensor gives you the following values across 10 exposures.

25419
25167
24957
25739
25601
25530
25104
24985
26010
25881

It's not wrong to average these, along with every other photosite. Sure, computing the new value is required and is technically computational but you're actually getting closer to the TRUTH of the real-world scene and not faking anything. You're just getting a cleaner sample.
When people talk about computational garbage they mean things like "enhancing" faces with AI or filters or applying tone curves and local contrast and other garbage like that.

In a perfect world our sensors would have larger wells and simply allow shooting at ISO1 if we wanted, making this workaround method unnecessary, but ultimately the result would basically be identical.

The only situation where a 600s (10m) exposure with liveND would be any less "authentic" than an actual 600s (10m) exposure with a physical ND is if the sensor readout and card dumping times are so damn slow you effectively "lose sight" of things long enough to cause problems, but that isn't really common. Without significant gaps between exposures you can do multi-sampling just fine without it adding anything artificial to the scene.

>>4446567
>applying tone curves and local contrast and other garbage like that.
Wtf, a tone curve is applied to every single raw file, it's not "computational" in the sense of AI.

>>4446591
No they aren't.
Data is data, the sensor data is a simple analog to digital conversion per photosite and you can demosaic or choose not to or apply the gamma correction for screen output or not to, RAWs are linear intensity data.
Adobe retards just assume that Adobe shit defaults or youtuber meme s curves are somehow inherent to photography but they aren't.