[stevebrot]

How am I supposed to use floodlights when I'm pursuing wildlife on a cloudy evening?

Works for poachers. It will work for you!


Steve

[DWS1]

Originally posted by luftfluss Quote
How am I supposed to use floodlights when I'm pursuing wildlife on a cloudy evening?


Read more at: K-3 III - Photons to Photos sensor DR measurements - Page 5 - PentaxForums.com

Works for poachers. It will work for you!


Steve

Just don't do it in a National Park. High dollar fine and an appearance before a federal judge.

[Mbaez]

Out of curiosity I shot a couple of pictures at ISO 25600 with my KP and I can see every whisker of my black cat in my bedroom with the lights off.. I guess Pextax noise cheating is really good and won't affect me in real world situations 😂😂😂.

[Ishpuini]

Pretty cool. I was just thinking how computational photography as in smart phones is becoming more acceptable, but it still is a taboo in dedicated cameras

Very true. Computational photography is the future, and regardless of the "early" version objections being theoretical or practical, it will improve from generation to generation. Sensors have only so much surface to catch whatever available photons that hit them during a limited exposure. Increasingly complex and well trained algorithms are probably the best way available to improve IQ beyond the physical limitations, and if these algorithms can take into account parameters that are not available to RAW processors because it is not possible to put these in the RAW files, then I am all for the application of these algorithms in-camera.

However, there is only so much possible, since the very concept of DSLRs only allows for one frame to be captured as a basis for the algorithms, which is completely different to what smartphones do. These basically don't capture any image when the shutter button is tapped, but instead will immediately read the last x images from their buffer and combine these into an image file. I.e. no shutter lag at all, and the resulting image will actually be mostly based on what the photographer last saw on his screen before tapping the button. The real power of smartphone computational photography lies in enhancing that exposure by combining it with information from previous images taken fractions of time before the base image (and all the photons from those multiple exposures). What DSLRs do only scratches the surface.

Wim

---

Last edited by Ishpuini; 04-30-2021 at 04:29 AM.

[wkadruje]

Out of curiosity I shot a couple of pictures at ISO 25600 with my KP and I can see every whisker of my black cat in my bedroom with the lights off.. I guess Pextax noise cheating is really good and won't affect me in real world situations 😂😂😂.

care to share the raw? I've never been _that_ brave and seen loss of detail at lower ISOs with my KP

[Mbaez]

care to share the raw? I've never been _that_ brave and seen loss of detail at lower ISOs with my KP

Sure, and don't be afraid: take a random picture at Hi-ISO and lets see together if the accelerator is really evil. It's an experiment worth doing instead of discussing without pictures. After all we have the cameras, so lets go and share pictures.

20210429_KPS_115.DNG - Google Drive
20210429_KPS_115.DNG - Google Drive

PS. The image looks way brighter than it was in real life.

[Kunzite]

the very concept of DSLRs only allows for one frame to be captured as a basis for the algorithms

That isn't true.

[wkadruje]

Sure, and don't be afraid: take a random picture at Hi-ISO and lets see together if the accelerator is really evil. It's an experiment worth doing instead of discussing without pictures. After all we have the cameras, so lets go and share pictures.

20210429_KPS_115.DNG - Google Drive
20210429_KPS_115.DNG - Google Drive

PS. The image looks way brighter than it was in real life.

thanks, will definitely compare that during the weekend to what I've been getting

[Ishpuini]

That isn't true.

Indeed, perhaps I was a bit too quick and brief with the statement that DSLRs one allow one frame as input for computation. Some technologies that are already implemented in our DSLRs and might be classified under computational photography, such as pixel shift, dark frame subtraction, in-camera HDR, ... do indeed expose multiple frames to produce one single image file resulting from some computation. All of which are taken *after* pressing the shutter release however.

Point is that the concept of DSLRs having a mirror between the lens and the sensor prior to the pressing of the shutter release, does limit the possibilities in a way that smartphones don't have to deal with. This means a DSLR cannot constantly refresh a buffer of a dozen or so exposures ready for processing at the press of a button, the way smart phones do it. I guess mirror-less cameras would/could be able to have such a buffer too? DSLRs with translucent mirrors or sophisticated AF sensors/meters might be able to collect some useful information before exposure, but not to the same level of detail. Perhaps in Liveview they could, but what would the added value of a DSLR compared to a mirror-less be then?

Regardless, there might have been some justified(?) resistance to this "baking" in the early stages, the technology noticeably not being sufficiently perfected yet, but I don't think we're far from no longer being able to detect any negative effects of computation in our images. Perhaps, we're already there in de K-3 III? I'll find out soon enough I hope... Keeping an open mind to technological advances...

Wim

[Kunzite]

Point is that the concept of DSLRs having a mirror between the lens and the sensor prior to the pressing of the shutter release, does limit the possibilities

I posted in a hurry thus I had to be (too) brief.
The concept of a SLR is much less limiting than you think. What you have in mind is the current configuration, with a primary and secondary mirror, dedicated AF system below and a metering system in the pentaprism housing.
A DSLR doesn't necessarily have to follow this - let's call it classic - configuration, though. Ignoring the technical hurdles, it could be simplified to a single, half-reflective mirror in front of an "always on" sensor. Said sensor would take care of metering and autofocus (just like on a mirrorless).
It would still be a DSLR, because it will be digital (obviously!), and it will have a single lens reflex viewfinder.

[Ishpuini]

I posted in a hurry thus I had to be (too) brief.

The concept of a SLR is much less limiting than you think. What you have in mind is the current configuration, with a primary and secondary mirror, dedicated AF system below and a metering system in the pentaprism housing.

A DSLR doesn't necessarily have to follow this - let's call it classic - configuration, though. Ignoring the technical hurdles, it could be simplified to a single, half-reflective mirror in front of an "always on" sensor. Said sensor would take care of metering and autofocus (just like on a mirrorless).

It would still be a DSLR, because it will be digital (obviously!), and it will have a single lens reflex viewfinder.

Agreed. Didn't Sony do that for some of their last DSLRs? Perhaps not with the same amount of computation that would be possible now.

Still, Ricoh isn't there yet, and seems to stick to the more classic approach?

[Serkevan]

Agreed. Didn't Sony do that for some of their last DSLRs? Perhaps not with the same amount of computation that would be possible now.

Still, Ricoh isn't there yet, and seems to stick to the more classic approach?

Something similar, but the Sony A-mount SLTs (Single-lens translucent) used the reflective "part" (I think the entire mirror is a pellicle?) of the mirror only for the PDAF module, the transparent part let light to the main sensor for metering and photo taking. Therefore, the viewfinder was electronic (and the mirror was fixed in place).

Typical DSLRs use the transparent part of the mirror, instead, for PDAF (via submirror). What Alex suggests is skipping this submirror, using instead the main sensor straight away, while keeping the optical viewfinder.


Another option for increasing the brightness of the OVF even more would be to upscale the RGB+IR sensor even further (1080p?) and use only that sensor for AF (it already handles the eye-detection in the K-3iii), even with OSPDAF. That would get rid of the submirror assembly, so the main mirror would be a bit lighter (more fps possible, less travel time/shock?), and the OVF image would be brighter as all the light would be reflected (and the RGB+IR sensor shouldn't take up more light than it already does, I guess). This of course would have the drawback of not allowing for the same computational magic with pre-buffered shots, but it would keep the sensor colder and avoid OSPDAF on it.

[Kunzite]

Didn't Sony do that for some of their last DSLRs?

Sony did two kludges:
- first, a "Live View" capable camera - but instead of using the main sensor, it used a secondary sensor in the pentamirror's housing. It was awful.
- then, the "SLT" kludge, which tried to be a "mirrorless" with a SLR AF module glued in. But these weren't SLRs.

Ricoh Imaging's idea (only an idea, so far) is to simplify the DSLRs while gaining advantages available now only to mirrorless. Even the light loss is optionally avoidable, by raising the mirror (just like now, but it's one mirror instead of two) in low light for actual shooting.
I like it.

[house]

The complexity involved in enabling mirrorless features on a dslr is quite high unless some technical breakthrough is made with the mirror. It's difficult to imagine a solution that doesn't sacrifice viewfinder quality/brightness.

My hunch is that dslr's are better off honing their advantages. The trajectory Pentax appears to be taking.

[Belnan]

2000$ camera just for HD resolution jpegs on social media ?

Welcome to the new world, I suspect people spend a lot more than that.