High Resolution Or Low Light Performance?

[Batdude]

Guys, there is no way the 40MP sensor is going to do miracles when it comes to high ISO performance. Is just physics. Is the XH2S doing miracles at 26MP? No, so why are some already guessing that the 40MP is going to have "improved" high ISO performance?

So in reality, who needs what? Do you need a lot of resolution to make prints of the size of the empire state building or for cropping birds heavily, or high ISO performance? You can't get both.

It looks like the trend is resolution, and this seems to keep going and going. I'm looking at the images from the 26MP sensor and darn it man, I don't need more than that. I actually need a sensor with top notch low light performance. I don't need more than 20MP that's the truth.

I wonder what Fuji is going to do with the XT5 and so forth but it looks like the camera industry has brain washed many making them think they "must have" that high resolution sensor. So the question to me is will Fuji ever make a low light top of the line low light performer? What's up with that?

And don't tell me there is software for that. I have no desire to add more extra steps and time wasted to my existing workflow which is in fact already time consuming. Some things are just starting not to make any sense sorry.

 

[nnowak]

First, there is no correlation between pixel count and noise levels. If you are looking at individual pixels, yes, the higher resolution sensor will have noiser pixels. When printed out at the same size, the higher resolution sensor will have the same noise performance. If anything, the higher resolution sensor actually has a slight advantage with modern noise reduction software.

Second, all modern image sensors are pretty much at the limit of physics. It is not 2003, and the days of massive gains in sensor image quality are long gone. Canon is the only company of late with noticeable gains in high ISO performance, and they are just finally catching up to, but not surpassing, Sony.

Third, Fuji can only buy whatever sensor some other manufacturer produces, and Fuji is already buying the best crop sensors on the market. No current Canon or Sony crop body can match the DR and high ISO capabilities of the Fuji bodies.

Real simple, a lower resolution sensor will not give you better image quality. If you need better high ISO performance, you need a camera with a larger sensor. There is no impending technology that will make a future crop sensor perform like current full frame sensor.

 

[vcxz]

This is discussed a lot in the Nikon forums since they have had bodies with different megapixels for a long time (d750/850, Z6/7).

The low light argument is true if you blow them up so the pixels are the same size. If you keep the total image size the same I think the difference goes away.

The extra megapixels are probably because technology has advanced where it can be done economically. It's probably similar to how phone storage minimum sizes have gone up from 32, 64, 128 gb even if you only ever need 32 gb.

Fuji's new philosophy is to offer great video specs in a cheaper and compact body. It has better video specs than stuff like the Sony AIV or Nikon Z6.

I just saw a really good video comparing the XH2 to Sony this morning:

So yeah if you're mostly into stills and ultimate resolution Fuji may not be #1. I was surprised the Nikon Z primes are cheaper than Fuji's new ones. But with the world going to video I can't blame them for making that their main focus.

 

[Batdude]

I fully agree with you and Batdude: IMOO the story is that camera makers have chosen to sell higher and higher resolution and the fastest possible autofocus. Personally, I have choses Fujifilm cameras for their lenses and the bodies size. This is because I usually take photos of people around me, and I need to go unnoticed. But - above all - I need to be able to shoot in dark environments (churches, houses, bars, nightplaces).

I frankly do not understand why Fujifilm had chosen to ignore the high iso problem. They could have made an X-T3S or X-T4S with a 20 mpixels sensor, allowing us to reach high sensitivities in darkness (over 6400 Iso): the same choice Sony made with their A7 S line.

Genuine question but is there another APS-C camera manufacturer out there who has better performing low light sensors?

FWIW, compared to any of the Canons I used to own (DSLR & Mirrorless) they were absolute garbage for both high ISO and dynamic range compared to what Fuji can do.

I said it above already, if low light/high ISO performance is your line in the sand as far as photography then just get a FF or even MF.

Precisely! There seems to be this assumption that Fujifilm is leaving a significant improvement in IQ on the table by not simply retooling and/or updating their sensors. So, let's stick with "apples to apples" comparisons here and please find another vendor who offers a crop sensor that delivers significantly better IQ. If such a thing is out there, then I think these demands for significant IQ improvements in next gen crop sensors are justified. If not, then my request to the OP is to please do us all a huge favor by getting off the pulpit, acquiring your choice of FF or MF camera, and just moving on.

Wise words as always, but common sense is not abundant around here... one would think people would choose the appropriate tool for their requirements.

I couldn’t agree more. I suppose it’s grumpy Mod time again.

“Choosing the appropriate tool” takes effort and costs money. Sharing one’s frustration with the fact that crop IQ simply can’t measure up to FF time and time again in the forum costs nothing. Yet, the solutions are, and always has been, right there for the taking.

1) Move to FF. Pick your vendor… there are numerous choices.

2) Go for the gold… move to MF. Even better IQ, but it comes at a price in terms of cost and portability.

3) If there’s a crop sensor body out there with better IQ out there, buy it.

If the OP truly believes that Fujifilm is leaving potential IQ on the table, then provide some proof in the form of a direct (and fair) comparison with another crop format camera. If that’s not possible (and I have my doubts), then either suck it up, pay the price (in cost, size, and weight), and move to FF or MF. If you can’t tell, the constant whining in post after post is just getting old.

--
Jerry-Astro
Fuji Forum co-Mod

Dr. Leonard McCoy would say “dammit Jerry, I’m just a damn photographer!”

Perhaps someone else is more capable than me to provide such proof as to why most likely a 40 MP sensor will perform worse in low light.

Again, I love my fujis which is the exact reason why I express my personal opinions what I feel about the use of my gear. I have the right to do so because this is a gear forum, and because I pay to use this gear out of my pocket.

As I have clearly stated, my concern is the future cameras which is being said they will have higher resolution sensors, which is something I don’t need in my work flow nor my photography application.

It would be nice if we had the option to be provided with a modern high quality lower resolution sensor designed for better low light performance. In many areas in the professional field photographers do NOT use nor want massive files, heck they don’t even shoot RAW for that matter.



You guys can shoot with whatever you want and I don’t have a problem if you buy high resolution sensors, the problem is why should some of us get stuck with gear that is not for our work application.



And yes I totally agree with you if nothing changes and Fuji will no longer work for my (photography) needs then yes I will most likely move to a different brand

 

[Eulan13]

I do agree. It does not make sense to reply me : "If you don't like it, change your system and leave us alone" as someone did. I have been shooting only with Fujifilm cameras since 2014 and this should make it clear that I like the cameras , the lenses, etc. THis is why I added a few of my images, just to show what my photography tends to be.

My complaint is that for marketing choices (but this is not only for Fujifilm, of course), the research is going towards speed and megapixels, without considering the possibility of making cameras REALLY different from mobile phones in terms of image quality.

Last but not least, I am an amateur photographer (even if with many personal exhibitions in Italy and abroad). So don't expect me to agree with those who easily suggest " Well, go to FF or MF". I know it perfectly, the problem at this stage is mainly financial. But, on the other end, am I allowed to say that I don't give a.. about 40mp or 30 frames/second but I do care about low light capabilities of a sensor? Am I offending someone in particular if I say that I do not agree with this marketing?

Thanks everybody for the technical information, always very useful.

 

[forest dream]

Real simple, a lower resolution sensor will not give you better image quality. If you need better high ISO performance, you need a camera with a larger sensor. There is no impending technology that will make a future crop sensor perform like current full frame sensor.

Future stacked sensor 2.0 tech - 2-Layer Transistor Pixel will improve DR and reduce noise.

Loading…

www.dpreview.com

 

[Batdude]

Exactly. OP could pick up a 20mp FF Canon if that’s all the resolution they require.

It's not that simple. You must not follow Batdude's never ending complaints against the cameras produced by just about all the camera companies out there.

You nailed it bro. I am not loyal to any brand and all camera companies out there have awesome pros but also major flaws that need to work on. So please don't get personally offended about your favorite camera brand because I'm expressing something negative that I don't like.

I can assure you, there's going to be multiple reasons put forth as to why a 20 MP FF Canon camera isn't an option. Near the top of the list will probably be the colors simply don't match the X-T1.

That is funny haha.

 

[robert1955]

I do agree. It does not make sense to reply me : "If you don't like it, change your system and leave us alone" as someone did.

I don't know who said that, but the reality is that FF will offer you one stop more [but also is at its technical physical limitations]

I have been shooting only with Fujifilm cameras since 2014 and this should make it clear that I like the cameras , the lenses, etc. THis is why I added a few of my images, just to show what my photography tends to be.

My complaint is that for marketing choices (but this is not only for Fujifilm, of course), the research is going towards speed and megapixels, without considering the possibility of making cameras REALLY different from mobile phones in terms of image quality.

Last but not least, I am an amateur photographer (even if with many personal exhibitions in Italy and abroad). So don't expect me to agree with those who easily suggest " Well, go to FF or MF". I know it perfectly, the problem at this stage is mainly financial. But, on the other end, am I allowed to say that I don't give a.. about 40mp or 30 frames/second but I do care about low light capabilities of a sensor? Am I offending someone in particular if I say that I do not agree with this marketing?

it is more than marketing, it is technical and physical limitations. Actually, the marketing people would love to market better lower light capabilities. I remember there was even a lot of noise about marketing pushing 'insane' iso numbers.

 

[Jerry-astro]

I do agree. It does not make sense to reply me : "If you don't like it, change your system and leave us alone" as someone did. I have been shooting only with Fujifilm cameras since 2014 and this should make it clear that I like the cameras , the lenses, etc. THis is why I added a few of my images, just to show what my photography tends to be.

My complaint is that for marketing choices (but this is not only for Fujifilm, of course), the research is going towards speed and megapixels, without considering the possibility of making cameras REALLY different from mobile phones in terms of image quality.

Last but not least, I am an amateur photographer (even if with many personal exhibitions in Italy and abroad). So don't expect me to agree with those who easily suggest " Well, go to FF or MF". I know it perfectly, the problem at this stage is mainly financial. But, on the other end, am I allowed to say that I don't give a.. about 40mp or 30 frames/second but I do care about low light capabilities of a sensor? Am I offending someone in particular if I say that I do not agree with this marketing?

Thanks everybody for the technical information, always very useful.

I think there are some pretty large assumptions that both you and Batdude are making: which is that Fujifilm has consciously embarked on an R&D strategy that primarily prioritizes MP over IQ. More importantly those assumptions also suggest that the sensor techology available today offers the ability for improvement both in density and IQ. That’s where I start having some concerns about the stance you’re taking. Developing next gen sensor technology with significant improvements in IQ likely presents a whole lot more of a challenge than one might think. Continuous density improvements (i.e. more MP) in process technology, however, have been demonstrated and implemented for years now in all varieties of CPUs, sensors, and other silicon. IQ improvements over that same period of time have been far less apparent, and my suspicion (and it’s just that — I’m no expert on this either) is that those improvements in IQ will come at a far slower pace and will require major modifications in both the underlying sensor technology and algorithms.

Bottom line: it’s mostly a matter of “your guess is as good as mine” here, but I’m not surprised that Fujifilm (and other manufacturers) are making major advances in density, but far less progress in ongoing IQ improvements. If I’m missing something here, please by all means set me straight.

--
Jerry-Astro
Fuji Forum co-Mod

 

[sluggy_warrior]

I fully agree with you and Batdude: IMOO the story is that camera makers have chosen to sell higher and higher resolution and the fastest possible autofocus. Personally, I have choses Fujifilm cameras for their lenses and the bodies size. This is because I usually take photos of people around me, and I need to go unnoticed. But - above all - I need to be able to shoot in dark environments (churches, houses, bars, nightplaces).

I frankly do not understand why Fujifilm had chosen to ignore the high iso problem. They could have made an X-T3S or X-T4S with a 20 mpixels sensor, allowing us to reach high sensitivities in darkness (over 6400 Iso): the same choice Sony made with their A7 S line.

Just so we are on the same page regarding high ISO performance, here's an example of ISO 12,800 (actually bumped +2.5EV in post, more like ISO 25,600) fed through a neural-network denoiser. Although this is just an early prototype trained with limited samples, personally this denoiser is more than I can ever asked for, totally worth adjusting my workflow for.

RAF: https://drive.google.com/file/d/1KfEy5_CZNa66_kd-qJi_dDaqCF-J572E/view


SOOC JPG


exported from darktable, then fed through nind-denoiser and RL-deblur

 

[forest dream]

My complaint is that for marketing choices (but this is not only for Fujifilm, of course), the research is going towards speed and megapixels, without considering the possibility of making cameras REALLY different from mobile phones in terms of image quality.

Current main difference between dedicated camera and smartphone is shooting experience :

• Ergonomics
• Manual exposure control
• EVF (framing under sunny day)
• Tilting screen

Agree that sensor size and optical glass play important on IQ, but many camera maker really not much innovation of IQ improvement.

IMO below area may improve IQ if camera maker focus on it :

• Better auto-ISO algorithm (always stick to dual native base-ISO if possible), let consumer switch SS dial within actual SS or min SS via physical lever.

• Built-in ND filter (either DJI Ronin 4D's switch ND glass or Electronic Variable ND filter)
• Dual native base-ISO, seamless integration with built-in ND to auto achieve base-ISO if possible
• Larger pixel via "Stacked Sensor 2.0 - 2-Layer Transistor Pixel"

https://dpreview.com/news/220084477...-its-new-2-layer-transistor-pixel-design-tech

• DJI LiDAR Range Finder - use LIDAR for assist AF especially low light

https://dpreview.com/news/946576490...ls-the-rs-3-dji-rs-3-pro-and-dji-transmission

• Built-in DeepPrime like denoising and demosaicing engine for improve noise performance before generate RAW file.
• Electromagnet shutter for minimize shutter shock

https://dpreview.com/news/742836567...er-design-that-aims-to-minimize-shutter-shock

 

[sluggy_warrior]

• Built-in DeepPrime like denoising and demosaicing engine for improve noise performance before generate RAW file.

Noooo! I'd prefer to have the RAW data as-is. The camera can apply fancy denoise/demosaic before exporting to SOOC JPG, but leave my RAW file alone. I'm sure the software on my computer will be way better than the one in the camera.

 

[forest dream]

• Built-in DeepPrime like denoising and demosaicing engine for improve noise performance before generate RAW file.

Noooo! I'd prefer to have the RAW data as-is. The camera can apply fancy denoise/demosaic before exporting to SOOC JPG, but leave my RAW file alone. I'm sure the software on my computer will be way better than the one in the camera.

Fujifilm X-T3 sensor (Sony IMX571 26MP BSI-CMOS sensor) seem like built-in 16-bit A/D converter. Are it mean the sensor can maximum output 16-bit data?

https://www.sony-semicon.co.jp/e/products/IS/camera/product.html

But, Fujifilm RAW - RAF only 14-bit RAW (uncompressed or lossless compression).

If camera built-in DeepPrime like denoising and demosaicing engine, are it possible direct denoising and demosaicing from 16-bit sensor data before oversampling to 14-bit RAW?

Are these built-in DeepPrime engine perform better IQ (via 16-bit data) than post-processing 14-bit RAW in PC?

 

[nnowak]

Real simple, a lower resolution sensor will not give you better image quality. If you need better high ISO performance, you need a camera with a larger sensor. There is no impending technology that will make a future crop sensor perform like current full frame sensor.

Future stacked sensor 2.0 tech - 2-Layer Transistor Pixel will improve DR and reduce noise.

https://www.dpreview.com/forums/post/66316051

Apparently you missed the part in the video where they state the technology was developed for use in smartphones. Because of the significantly smaller sensors, and even smaller pixels, technologies that are meaningful for a smartphone sensor have very little benefit on the much larger sensors used in interchangeable lens cameras. BSI was originally developed for smartphones where it was quite beneficial. The X-T3 sensor is BSI and it has no perceptible improvement in image quality over the the standard CMOS sensor in the X-T2.

 

[forest dream]

Real simple, a lower resolution sensor will not give you better image quality. If you need better high ISO performance, you need a camera with a larger sensor. There is no impending technology that will make a future crop sensor perform like current full frame sensor.

Future stacked sensor 2.0 tech - 2-Layer Transistor Pixel will improve DR and reduce noise.

https://www.dpreview.com/forums/post/66316051

Apparently you missed the part in the video where they state the technology was developed for use in smartphones.

Both BSI-CMOS & stacked sensor also used smartphone too.

Smartphone sensor use 2-Layer Transistor Pixel to increase light gathering in small pixel.

• Sony a7IV (33MP FF BSI-CMOS sensor) pixel size = 5.93µm.
• Fujifilm X-H2 (40MP APS-C BSI-CMOS sensor) pixel size = 3.03µm
• OM System OM-1 (20MP MFT stacked BSI-CMOS sensor) pixel size = 3.36µm

X-H2 pixel size only around 51% of Sony a7IV (FF). It is mean reasons why FF's low light performance better than APS-C.

Because of the significantly smaller sensors, and even smaller pixels, technologies that are meaningful for a smartphone sensor have very little benefit on the much larger sensors used in interchangeable lens cameras.

In FF point of view, 40MP APS-C's pixel size consider small (only 51% when compare with Sony a7IV).

BSI was originally developed for smartphones where it was quite beneficial.

Below article mentioned 2-Layer Transistor Pixel which "separates the photodiodes and pixel transistors into two different substrate layers. By splitting the layers, the photodiode can be larger, resulting in a higher signal saturation level (also referred to as full well capacity), while the larger transistors result in less noise."

https://dpreview.com/news/220084477...-its-new-2-layer-transistor-pixel-design-tech

It means 2-Layer Transistor Pixela able make APS-C photodiode larger and sensor less noise, more close to FF performance.

The X-T3 sensor is BSI and it has no perceptible improvement in image quality over the the standard CMOS sensor in the X-T2.

• No all consumer require BSI-CMOS.
• No all consumer require Stacked BSI-CMOS.
• No all consumer require 2-Layer Transistor Pixel tech.

For consumer feel that X-T2 (FSI-CMOS) is good enough, can continue stick on X-T2, no need purchase X-T3/X-T4...

For consumer reject BSI-CMOS but demand better AF performance, may be can consider Canon R10/R7.

P.s. Canon R10/R7 still recycle old FSI-CMOS sensor since ...

 

[biza43]

As already stated:

1. Higher mp count does not reduce high ISO IQ, if you print at same size.

2. You still do not say what limits Fuji are imposing on your workflow in terms of high ISO. What is your ceiling? What is the ISO above which Fuji stops working for you?

3. Only increase in sensor area will improve high ISO IQ. Look at #2, and then decide whether FF or MF might be better for you. An APSC sensor at 20 mp will not give you better high ISO IQ compared to a 40 mp APSC sensor, if you keep print size the same.

 

[robert1955]

I do agree. It does not make sense to reply me : "If you don't like it, change your system and leave us alone" as someone did. I have been shooting only with Fujifilm cameras since 2014 and this should make it clear that I like the cameras , the lenses, etc. THis is why I added a few of my images, just to show what my photography tends to be.

My complaint is that for marketing choices (but this is not only for Fujifilm, of course), the research is going towards speed and megapixels, without considering the possibility of making cameras REALLY different from mobile phones in terms of image quality.

Last but not least, I am an amateur photographer (even if with many personal exhibitions in Italy and abroad). So don't expect me to agree with those who easily suggest " Well, go to FF or MF". I know it perfectly, the problem at this stage is mainly financial. But, on the other end, am I allowed to say that I don't give a.. about 40mp or 30 frames/second but I do care about low light capabilities of a sensor? Am I offending someone in particular if I say that I do not agree with this marketing?

Thanks everybody for the technical information, always very useful.

I think there are some pretty large assumptions that both you and Batdude are making: which is that Fujifilm has consciously embarked on an R&D strategy that primarily prioritizes MP over IQ. More importantly those assumptions also suggest that the sensor techology available today offers the ability for improvement both in density and IQ. That’s where I start having some concerns about the stance you’re taking. Developing next gen sensor technology with significant improvements in IQ likely presents a whole lot more of a challenge than one might think. Continuous density improvements (i.e. more MP) in process technology, however, have been demonstrated and implemented for years now in all varieties of CPUs, sensors, and other silicon. IQ improvements over that same period of time have been far less apparent, and my suspicion (and it’s just that — I’m no expert on this either) is that those improvements in IQ will come at a far slower pace and will require major modifications in both the underlying sensor technology and algorithms.

Bottom line: it’s mostly a matter of “your guess is as good as mine” here, but I’m not surprised that Fujifilm (and other manufacturers) are making major advances in density, but far less progress in ongoing IQ improvements. If I’m missing something here, please by all means set me straight.

Adding to this: Fuji buys its bare sensors from Sony Semiconductor. Neither that supplier nor any of the others seems to be able to make significant progress in low light capabilities

 

[forest dream]

Adding to this: Fuji buys its bare sensors from Sony Semiconductor. Neither that supplier nor any of the others seems to be able to make significant progress in low light capabilities

Fujifilm X-T3 sensor (Sony IMX571 26MP BSI-CMOS sensor) built-in 16-bit A/D converter, which maximum output 16-bit sensor data.

https://www.sony-semicon.co.jp/e/products/IS/camera/product.html

But, Fujifilm RAW - RAF only 14-bit (uncompressed or lossless compression).

Are it means some data loss when X-Processor 4 convert sensor data into RAW?

If Fuji firmware enough good, are it possible analysis 16-bit sensor data first (clear some noise) before convert into final 14-bit RAF RAW file?

 

[forest dream]

As already stated:

1. Higher mp count does not reduce high ISO IQ, if you print at same size.

Below article mentioned "...By splitting the layers, the photodiode can be larger, resulting in a higher signal saturation level (also referred to as full well capacity), while the larger transistors result in less noise."

Loading…

dpreview.com

In other words, the article mentioned:

• Larger photodiode (in each pixel)- more full well capacity photodiode.
• Larger transistors (in each pixel) - less noise.
• And usually larger pixel has larger photodiode and larger transistors.
• Larger pixel size = less noise per pixel

 

[Batdude]

Adding to this: Fuji buys its bare sensors from Sony Semiconductor. Neither that supplier nor any of the others seems to be able to make significant progress in low light capabilities

Fujifilm X-T3 sensor (Sony IMX571 26MP BSI-CMOS sensor) built-in 16-bit A/D converter, which maximum output 16-bit sensor data.

https://www.sony-semicon.co.jp/e/products/IS/camera/product.html

But, Fujifilm RAW - RAF only 14-bit (uncompressed or lossless compression).

Are it means some data loss when X-Processor 4 convert sensor data into RAW?

If Fuji firmware enough good, are it possible analysis 16-bit sensor data first (clear some noise) before convert into final 14-bit RAF RAW file?

I don’t know much about this, Fuji rumors said this under the third paragraph: “It looks like Fujifilm did not/not yet enable 16bit readout mode, and went for 14 bit instead, because they prioritize speed over image quality.”

So (if) this is true, then my theory is correct and Fujifilm is simply following a trend and is trying to catch up to the competition by concentrating on speed and higher resolution and (not) image quality which includes better low light performance.

So I would say there is your proof. Disappointing if this was a fact but there it is. I personally rather pay for 16 bit over useless high FPS and a ton more megapixels that I simply don’t use.

Loading…

www.fujirumors.com

 

[tko]

Guys, there is no way the 40MP sensor is going to do miracles when it comes to high ISO performance. Is just physics.

It isn't physics. Remember that high MP sensor don't really have any more underlying noise than low MP sensor. However, a low MP sensor does act to "filter" the noise so you don't see it as much. You can always take a high resolution sensor and filter it to give better or equal performance than a low resolution sensor.

SO it's not the number of pixels that's the issue, it's that high resolution reveals more details -- and more noise. Since you can filter in post, a high resolution sensor gives you both.

And don't tell me filtering in post is too much trouble. Very few photos really need the low light capability of modern cameras. Filtering the few noisy ones only requires the push of a button.

What you can't do it get more resolution when you need it.