When testing cameras, it is not unusual to see a situation when one camera can produce results a bit darker or brighter than another. In some cases, lenses are to blame for this variance, since most lenses cannot ideally transmit all of the incoming light. What this means, is that a lens with a maximum aperture of f/2.8 could potentially transmit less light, which could be equivalent to say f/3.5 in terms of brightness. The latter number is what is often referred to as a “T-stop”, or Transmission-stop, which is basically an adjusted f-stop that takes into account this light loss. In other cases, the camera itself can be the source of brightness variance. Although manufacturers are supposed to adhere to an ISO standard that guides the process of determining the right brightness level for each ISO, there is usually still some variance between not only brands, but also between specific camera models. We won’t get into the question of why there are such variances. Instead, we will concentrate on implications of such variances to camera sensor comparisons and ratings. Particularly, we will be looking at exposure variances in Fuji cameras, such as the Fuji X-T1. Many photographers, including myself, have been fond of the way Fuji sensors render images, outputting very clean and pleasant-looking images, even at high ISOs. But are those ISOs real? And is Fuji doing something shady to make its images look better? Let’s take a closer look…
When testing cameras, I am pretty open about showing what camera settings I use for a particular scene. Knowing that light can have a huge impact on the scene and also knowing that the same source of light is practically never perfectly constant in brightness, particularly over a long period of time, I always retest every camera that I provide in comparisons. This way, you are not looking at a scene captured a year ago from one camera and the same scene captured a week ago from another camera. This process takes a long time, but I don’t mind doing it, as I want to only show our readers a more accurate representation of true sensor performance. Due to the above-mentioned variances, making a fair comparison between cameras proved to be rather difficult. How do we compare cameras? Do we compare them at the same ISOs at identical exposure levels? Do we compare them at the same ISOs, but at different exposure levels to match the same brightness? Do we compare them at the same brightness at equivalent ISOs? Or perhaps we should compare them at the same ISOs, but make further adjustments in post-processing to match brightness levels? As you can see, making the proper decision on comparisons is no easy task. So far, I have been comparing sensors based on the same ISO level, but making exposure adjustments to match the same brightness across different cameras. While this method works fine to show differences in noise levels at the same ISOs, differences in exposure can yield to potential issues for such comparisons, as we are not necessarily looking at an apples-to-apples comparison, particularly when variances in exposure are very high.
Fuji X-T1 vs Sony A7 II
Let’s take a look at a rather typical situation involving a Fuji camera (in this case the Fuji X-T1) and the Sony A7 II (Left: Fuji X-T1, Right: Sony A7 II):
The Fuji X-T1 had the Fujinon 35mm f/1.4 lens mounted on it, while the Sony A7 II had the 24-70mm f/4 lens mounted on it. Granted the two might not be equivalent in light transmission, for now we are simply looking at equivalent brightness between these two cameras. At the same exposure settings of ISO 200, f/8 and 2 seconds, it is pretty clear here that the Fuji X-T1 on the left looks significantly darker than the image from the Sony A7 II on the right. This difference in brightness makes it difficult to compare the two cameras properly, as one simply shows no shadow detail.
Let’s take a look at what happens if we compare the two cameras at the same ISO and exposure settings, but after making exposure adjustments in post to see how much variance there is between the two. Here is the same image from Fuji X-T1, but this time adjusted by +0.85 EV in Lightroom:
Now the two look pretty comparable. I tried increasing the exposure on the Fuji X-T1 to higher values, but it resulted in over-exposure. Looking at the JPEG histogram in FastRawViewer and a comparable RAW histogram with exposure adjustments indicated that the above change was more or less accurate in most regions of the image. According to my estimates, the real difference in exposure between these two cameras is roughly 2/3 of a stop – the rest of the difference comes from lens light transmission levels. A similar conclusion can be made from looking at the DPReview Studio Test Scene. I downloaded RAW files from both cameras and compared them in Lightroom with 0.66 EV adjustment on Fuji X-T1. Here are two images compared at ISO 3200:
This result is interesting, because it shows that despite the 0.66 EV boost that I gave to the Fuji X-T1, its images at ISO 3200 actually still look a little cleaner than on the Sony A7 II (and that’s with Sony A7 II images down-sampled to 16 MP). If you look at the files, you can clearly see that the Sony A7 II contains more chroma noise than the X-T1. This essentially shows that even if we were to equalize the X-T1 sensor performance, it still can be a challenge for the A7 II in IQ at high ISOs – this confirms my findings from my Sony A7 II review, which as I have demonstrated, cannot match the Nikon D750’s sensor performance either, although the D750 also sports a 24 MP sensor. But we are not here to discuss the A7 II – we are here to see if Fuji has been cheating with their cameras when it comes to real ISO performance.
Fuji X-T1 vs Nikon D7100
Let’s take a look at another example and compare the Fuji X-T1 with another APS-C camera, the Nikon D7100. Here are images from both at ISO 3200, the X-T1 is again adjusted by +0.85 EV in Lightroom:
Although the X-T1 lacks the details of the D7100 (down-sampling from 24 MP to 16 MP did its job here), the image looks slightly cleaner in comparison. Again, this is because the X-T1 image is showing less chroma noise than the D7100. This is an interesting result, because we essentially have somewhat of an equivalent comparison here, considering that the X-T1 image saw a +0.85 boost in post.
Fuji Real ISO Comparison
Now that we know that Fuji is cheating by underexposing its images somewhere around 2/3 of a stop, let’s take a look at how the actual ISO of the Fuji cameras look like compared to the “simulated” ISOs. Here is a comparison of real ISO 3200 vs Fuji’s underexposed ISO 3200:
As expected, the real ISO 3200 does look noticeably worse than the simulated ISO – the 2/3 of a stop difference that I adjusted in Lightroom shows more noise. Here is ISO 6400 comparison:
Again, it is clear that the results are different – ISO 6400 with 0.66 EV dialed in Lightroom looks comparably worse, showing more chroma noise and introducing some artifacts to the image, particularly in the shadow area.
Conclusion
By now, it is a pretty known fact among the photography community that Fuji underexposes its images by around 2/3 of a stop to a full stop when compared to other cameras when shooting RAW and using most commercial RAW converters. There are all kinds of theories out there, with some people claiming that Fuji does this on purpose to look better, while others attribute this difference to Fuji’s X-Trans sensor and the demosaicing process. I have to say, that despite my attempts to normalize the RAW files by making adjustments, Fuji still looked pretty darn good compared to other cameras. The main source of this is reduced chroma noise, which is evident when you look at RAW files from both the above comparisons and other sources, such as DPReview’s studio comparison tool. That’s the main reason why Fuji looks so clean for an APS-C sensor. I do not know exactly how Fuji achieves this, but the X-Trans sensor and its demosaicing process are probably the reason for the reduced chroma noise we see in images.
Update: Thanks to some awesome people like Iliah Borg, we now know the reason why Fuji RAW files appear darker. Turns out that Fuji has a special tag (0x9650) in its RAW files that highlights the necessary midpoint compensation for RAW files to interpret and make necessary changes. Below are the values for the Fuji X-T1:
ISO 200 / 0.72EV
ISO 400 / 0.72EV
ISO 800 / 0.72EV
ISO 1600 / 0.72EV
ISO 3200 / 1.38EV
ISO 6400 / 2.38EV
So keep the above in mind when looking at Fuji RAW files and comparing them to other cameras. If you are using a RAW converter from Adobe (and potentially other RAW converters), make sure to look at the above table for adjustments needed to make images appear as they should. Big thanks to Iliah Borg and the LibRaw team for discovering the Fuji tags and letting us know!
Interesting article.
I’m not sure I understand the “Real ISO Comparison” though.
Eric, basically, I showed what Fuji outputs at ISO 3200 and 6400 and what it really looks like when compared to other cameras at the same ISO…
I find your article quite interesting, but I also don’t understand how you created the images in the section “Real ISO Comparison”. It might help, if you could give more details about the camera settings and software / software settings you used to create the images.
I’m also not sure about this one:
Does Lightroom handle the raw exposure bias tag correctly, or do I have to correct the lighting in X-trans images manually in Lightroom?
Real ISO seems to me ISO6400 current from Fuji and then when you remove the “underexposition” and boost the ISO to match identical “brightness” of the picture. Then you have the current vs what it should be…
Hi Nasim,
Nasim, what a great attention grabbing title you have come up with!! Fuji Cheats!!!
Its been over one year now since your promise to review Fuji lenses……….Do you still intend to do so?
Wandiba, not sure if grabbing attention was the purpose – just wanted to illustrate with examples some facts behind Fuji’s ISOs.
As for lenses, I have all the data I need and MTF numbers to show. Just need more time to write the actual reviews :) Sorry it is taking so long – last year was a big challenge to keep up with everything that was going on with my life. This year I have some big plans, so please stay tuned :) Right now I am measuring MTF for the Fuji 16-55mm f/2.8. I might have a lemon, because I cannot get MTF numbers that are as good as on the XC16-50mm, which is just not right. The biggest screw-up on this lens that I see so far is the darn hood – it takes so much effort to put it on and remove it. I have a suspicion that I have actually damaged the lens while trying to mount and dismount the hood! So idiotic, I don’t know what a heck Fuji was thinking. The lens feels solid and heavy otherwise. If I cannot get any good results with it, I will have to test another sample later on…
You promised us reviews for Fuji’s primes lenses : 35mm, 23mm ,14mm and 56mm f1.2 …I remember these were your words when you reviewed Fuji X-T1. However it looks like we going to see 16-55mm f/2.8 review first before any of the afore mentioned prime lenses.
I delivered the 35mm f/1.4 review, but the others went into the queue. I apologize for that. I will do my best to prioritize these – life has been hectic and I am doing my best to manage time better.
Maybe these two could cover you financially while you stop doing your job and write those reviews you promised. As for me I’m still looking forward to the reviews if and when you get a chance to do them, in the mean time I’ll continue using those primes as I’m sure I’ll continue to be as happy with them even after I read your findings. ;)
Nasim, do the reviews as and when you like! :D
Would seem a fair test to do. Nice to see the facts made clear.
Nasim,
Thanks for the info/warning.
It seems incredible that a respectable manufacturer can do that… What do they expect, with all the tests on the Internet? Do they really think they could get away with it unnoticed?
However, you dig out the facts, so we’ve got to live with them.
Jean, who knows, perhaps the X-Trans sensor is to blame for these variances. The fact is, images from Fuji cameras typically come out darker, which is why their real ISO performance is a bit different. However, as you have seen from the above comparisons, even if Fuji did not cheat, their results would still look pretty solid. So I am not sure why they did what they did…
“It seems incredible that a respectable manufacturer can do that…” if by ‘that’ you mean use a alternative yet entirely legitimate definition of ISO which is well within the standard then yes, how dare they not conform with what everybody else is doing, the horror of even developing an unorthodox colour array!! the horror!!
Easy go, Sam, after the discussion in the comments that shed some light on the question. :-)
Jean, your initial comment supports my concern about how misleading (or misrepresentative) using such loose comments like ‘cheating’ by people of Nasim’s standing on the site can have on readers. Granted, the article does go on to comment about how this is not necessarily a bad thing (as it protects highlights and can be pushed up in post) but that’s not enough to dispel the myth that Fuji is somehow not playing by the “ISO rules”
Sam, I was not the first person to talk about Fuji’s misleading ISO numbers. I have seen these issues in the past and wanted to bring this up to our readers’ attention, particularly after being accused of being a Fuji fanboy, for providing incorrect data or comparisons. So the point of the article was to address such concerns and to explain, that EVEN if Fuji cheats, the results at comparable ISOs still place Fuji ahead of most competition. I did this to show the reality and I am thankful that I did, because people like Iliah Borg showed exactly what’s going on behind the scenes. So we all learned something from this and now there is reference material that we can actually refer to when “fanboy” comments pile up.
Nasim, I understand what you were trying to do and why, and you did manage to get the point across, there’s no argument from me in that context. My comments have been very specific and refer exclusively to the repeated use words to the effect of ‘Fuji cheats’ throughout your comments. I think that once you have (successfully) accomplished the task of clarifying that even if their untouched files are different to others with the same settings, once compensated they produce similar and most of the time superior images (noise-wise) it’s time to stop ‘over-compensating’ (pardon the pun) for other’s perceptions of where you stand with Fuji products. Fanboy or not fact is Fuji does not cheat, they set a different midpoint on their RAW files and thus choose to use an alternative (yet still valid) definition of output well within the specifications. There is nothing technically misleading in their approach and thus the label ‘cheat’ is misrepresentative and misleading when used repeatedly by someone of your standing.
Let me stress that I don’t have an issue at all with the article, and I’ll be the first person to defend your right to publish anything you choose on your site (in fact I’ve done a number of times in relation to phones OS reviews etc ;) ) and I do appreciate all the work and effort that goes into preparing and researching your posts.
Sam, since that word seems to have offended you, I went ahead and edited my comments and removed it. Hope you won’t get offended anymore. Didn’t realize the word was so negative to some, but perhaps it is me who should avoid using them in the first place, so please forgive me.
Nasim, I think the issue is not whether I’m offended by them or not. I appreciate the gesture, honestly, and maybe there is a slight but evidently significant difference in the connotation of the word to us here in Australia where a “cheat” is someone (in this case something) not worth of our time or attention as they are beneath us ‘honest’ lot.
I guess this is another word which has a far more negative and derogatory significance in places like Australia than it does in the US. For example, I watched in disbelief a US swimming champion refer to a pre-swim warm-up gesture as ‘looking like a spastic’ to which people from across the Atlantic and Australia jumped in criticism due to the un-pc use of the term. US comments of course defended her in that the word does not mean the same in the US as it does everywhere else in the English speaking world. I believe that if we are publishing this material to an international audience we should really pay more attention to how we word our opinions.
There’s nothing to apologise for and I’m sorry you felt the need to change your comments because as illogical (maybe) as it sounds you do have every right to say whatever you want just like I have every right to criticise you for saying them.
I should also say that I’m perhaps more critical of your opinions because they are ‘your’ opinons, which I deeply respect even when I sometimes may disagree with.
Well, Sam, shall we go as far as to say that Nasim may have yielded to the temptation of a more attention-grabbing title than necessary, given the popularity of PL? Probably…
Actually I’m into photography and what I see rather than such technicalities, but I was interested in what Nasim had noticed… For all that really matters, I don’t think my pics are any better now than they used to be in the film age — the biggest difference being, in my humble opinion, in the price of each photo: we HAD to carefully prepare for each photo when what we had inside the camera was a roll of film and only 36 possibilities, and not a card accommodating hundreds of more or less failed attempts… ;-)
Hi Jean, nothing wrong with seeking attention and using headings as hooks particularly in this instance as it’s a question rather than a statement. :)
I’m with you on the point about film, I used a composition card (similar to the ones in this article photographylife.com/const…ualization ) for years so I wouldn’t be tempted to press my shutter unless I was totally convince the shot was going to work. ;)
Of course, every ill-pressed shutter had to be paid for! And not rarely do I still treat my DSLR as if it were an SLR. Btw, lately I found a nice and fun app for smartphone. It’s called “1-Hour Photo” and it delays the viewing of the snapshots you’ve taken with your iPhone (only iOS, I’m afraid) by one hour. I liked the idea, bought the one-dollar app, but haven’t used it yet… ;-) You may want to have a look at the company’s reasons for developing that app — check nevercenterlabs.
Ha ha! believe it or not I have in the past covered my LCD with cardboard and tape around the edges to simulate the ‘wait and see’ process as well. I’m fairly sure the keeper rate for those compositions would be noticeably higher than my normal shooting approach.
Of course, I believe you, Sam—although it’s not far from calling for a shrink… :-) But don’t worry, everybody is someone else’s fool, especially me…
Jean, for my your comment here has a great deal of merit. Having undergraduate and graduate degrees in photography I did have to learn the technical details for, at that time, film. However, after shooting for 50 years I’m not so interested in anything other than producing images for a client. As it relates to this thread, were I to obtain a Fuji camera I would be doing practical testing to understand the equipment. Once done, I really wouldn’t matter to me if Fuji was a stop off its ISO or not. I would just know that it performs in a certain way as opposed to other equipment and make my adjustments as needed. I think that is what you are implying and I agree with you totally.
Yes, Mike, although not my prime concern, I also imply what you say.
After all, when we bought a 50-ASA or a 400-ASA roll of film/transparencies, we knew we would have to adapt until the last picture of the roll was exposed, whatever the weather or brightness or darkness of our subjects.
But I would not like to sound too veteran-like—the liberty and flexibility the adjustable ISO of our modern DSLRs gives us is invaluable. And the advantage we have over the young guns is that it doesn’t go without saying and we know how to appreciate it… :-)
Jean, again very true.
The cost of individual frames is of course vastly less than with film, but I am not so sure when we factor in the cost of all the equipment needed to acquire, process, store and display or print all that digital ‘stuff’, whether overall it’s not a lot more costly than it used to be with film?
Yes, Betty, It would be costly indeed, should one specifically purchase all the equipment needed for photography. Most people have a computer though, and if they have to add a 1-TB hard drive, nowadays such a disk costs very little, surely much less than a year of film use.
As I work in the publication business, I would have the same computers, SSDs and hard disks even if I were not into photography. So I can say it costs me close to nothing compared to film. :-)
Betty, a true cost analysis would bear out your hypothesis but I think you may be correct. Although, back in the film days, I never did my own color, my black and white dark room cost a lot to build and equip. (We should bare in mind that the dollars are different now but could be equated.) There was still storage for negatives and slides not unlike the back up drives we use today on they were analog not digital. I think the only difference really is shooting a lot more at one time of any subject because it is very easy to toss out the duds and just keep what we think are the keepers. When I do still life for myself, I still calculate my exposure and light the same way I did in the film days, but when second shooter at a wedding, I can produce 1500 to 2500 images in a full day affair. I’ve fallen victim to the digital exposure process.
Yes, overall for me digital is infinitely more flexible and allows one to experiment and learn from mistakes cheaply and quickly. Discovering that one had made an exposure error 10 days after a shoot with Kodachrome was such a pain!
Yes, for me digital is so much more flexible and allows one to experiment and make mistakes relatively cheaply.
It used to be such a pain discovering an exposure error 10 days after a shoot using Kodachrome!
I think flexible is a good way to see the difference between film and digital. Even post processing can be called flexible as opposed to film.
Wow Sam, your comments come over as pretty aggressive, if I may say so! Maybe you should take things a bit more lightly?
OMG! The horror, the horror.
Guess I better sell my XT-1 and erase all those great looking photos I took with it.
Steve Z
Steve, of course not! Did you read the article? :) My X-T1 is a keeper, even though Fuji cheats.
Of course he did not bother to read the whole article, some people just go straight to comments, fanboys do that a lot.
Incisive. Succinct. Devastating.
To the batmobile!
This has been known for some years now. Unfortunately some “review” sites still continue to neglect this fact. Fuji XE2’s ISO3200 takes same amount of light as NEX-5T’s ISO1600, so when you compare ISO3200 images XE2 looks cleaner but actually XE2 RAW is at ISO1600 instead of 3200.
On the other hand, XE2’s RAWs at ISO6400 are as good as NEX-5T RAWs at ISO3200, so no problem there as long as you know what you are doing.
I consider Fuji’s X cameras as ISO100-3200 cameras, not 200-6400 as advertised.
If you set the ‘Dynamic Range’ option to 100% in the menus you will find this behaviour goes away and the reported ISO in the RAW file becomes the true ISO.
What is happening is that if you have DR set to anything else it underexposes so that the resultant JPEG (either directly or from software that processes the RAW in to JPEG afterwards) does not blow the highlights. Not a bad thing to do, but it’s pretty stupid that the RAW file does not report the true ISO.
Robert, Dynamic Range is always set to 100% on my X-T1, so it is certainly not the case…
Are you sure Nasim?
A friend of mine has an X100 and reported this problem. I googled it, found many posts stating this occurs if DR is not at 100%, so I made the change on his camera and it was resolved.
Yes sir, absolutely positive. I knew about the DR issue long ago, so I have been using 100% ever since…
I think it’s best than to dial in +2/3 exposure comp in most cases to get the correct exposure and to prevent additional noise from increasing exposure in post-prod. For example, i usually slightly underexpose shots with my nikon to save details in the highlights, because I know shadow detail is very recoverable on a Nikon.
Daniel, dialing +2/3 of exposure compensation won’t do anything, because you are just going to end up blowing out the highlights. The Fuji X-T1 exposes properly, that’s not the issue – the issue is when you compare the same exposure across different cameras.
Also, underexposing to save details in the highlights is not a good technique – you actually want to always expose to the right to get the shadow detail. Underexposing is unnecessarily losing data…
Nasim, isn’t it 4am in your part of the country? Why aren’t you sleeping? Thanks for the tip about Nikon, I’m always concerned with clipping the highlights, so i thought underexposing by -.3 to -.7 would help that. I’m talking about high contrast lighting situations. If the light is even, i’d not dial in any additional compensation.
Daniel, it is actually 5 AM. I am very tired, but I have to finish testing the lenses right now. Tomorrow I am leaving to San Francisco and I really need the data to be able to work while I am there.
Underexposing can be good if the lighting situation is complex and you are blowing out a lot of data. In all other cases, it is best to overexpose a little – that preserves the most amount of data in photos.
Thank you for your commitment, dear sir. I admire your dedication.
Thank you for the advice.
Not being argumentative, but if data is getting blown out, then overexposure is taking place.
And is ETTR overexposing or exposing correctly – to gather maximum data without clipping?
Betty, blown out means different things on different mediums. If part of an image is getting blown out in the camera, it might not be blown out in RAW at all, since you are still looking at a rendered JPEG when looking at the camera LCD. A little bit of overexposure is usually OK in the camera – it still means you can recover most of the data in post, since you will be pulling the data from a RAW file. That’s why ETTR usually works pretty well – since you are pushing to the right in the camera.
But if you are overexposing on the RAW file, then the data is gone. That’s why it is a good idea to look at how each camera captures images and understand differences in what you see in camera histogram vs RAW histogram. That’s why I love software like RAWDigger and FastRawViewer, because I can assess the camera histogram and compare it to the RAW histogram I see in software. This way, I can understand how much I can push to the right on the camera. Really wish manufacturers implemented RAW histograms – would make our lives so much easier!
Yes, understood. To me blown out means clipped in the RAW file and therefore unrecoverable.
Yes, also totally agree, if manufacturers gave us a RAW histogram it would save a lot of ‘guesstimation’.
It would be better still if they if they calibrated camera meters to indicate correct exposure just short of white point clipping (instead of outdated 18%grey) for photographers who know what they are doing and why they are doing it!
Maybe a ‘pro meter switch”?
Betty, write a letter. That would be an extra $500 I would be willing to spend in a new camera body. Don’t think it will ever happen because folks like you will compensate naturally based upon experience. New photographers will learn to do the same by reading a quality blog like PL.
Isn’t that what Nikon’s Highlight-Weighted Metering does in the D810 and D750?
+1 I’d also prefer to have a live RGB histogram while they are at it.
I wonder if the new “Natural Live View” setting included in a recent Fuji update helps to deliver a more accurate histogram – one that does not have the shadows/highlight effects of the JPEG film simulations and more closely resembles the RAW file histogram.
And to answer the question, ETTR is exposing correctly to gather maximum data, without clipping anything (well, clipping the brightest parts of a white cloud is OK with me, but typically nothing else). That’s how ETTR should be ideally :)
Now I’m thoroughly lost. On my X T1 I find that it overexposes ⅔ of a stop if I shoot in matrix metering, and -⅔ of a stop if I use average. So what is going on here? As to ETTR, is it not easier to recover shadows in post than highlights?
What is going on is you are using two different metering methods.
If you used a handheld incident meter you would get a third reading different from the two you already have. And if you used a spot meter anything could happen depending on which tone you pointed it at.
The point is a meter is a guide which has to be intelligently interpreted, not followed blindly.
As to ETTR, if either the highlights or the shadows are truly clipped, they are gone forever.
If the shadows are underexposed such that some data is left in one or more channels, then your RAW converter algorithm can interpolate the missing information and give a usable result. The problem is that at the shadow end of the histogram you may have only 8-16 tones to play with so banding and other artefacts will readily become apparent and noise will quickly rear its ugly head.
If the highlights are overexposed (but not completely), then again recovery is possible, but as fully 50% of ALL the data in the RAW file is held in that uppermost stop of exposure, there is a great deal of latitude and little risk of problems arising during further editing.
So, in a way, it is actually ‘easier’ to recover highlights than shadows.
But at the end of the day, there is no substitute for getting it right in the first place!
I am aware that the different meters produce different results. That is why I brought it up. I was wondering what metering was used for the test. since I generally use the matrix. As to ETTR, I find that if I nail exposure and are more careful not to clip the highlights there is always enough shadow info to produce my desired results. But thank you for your input. At the end of the day, I guess we simply have to know the camera.
Thanks
Hi Daniel
You wrote “dial in +2/3 exposure”. This means overexposing by 2/3. But in this way you would blown the highlights, though you don’t want to do it. I hope you see the source of this misunderstanding.
Best regards
Yes, Carmelo. Nasim also pointed this out, too. Thank you.
Interesting article Nasim. I am in a situation where having a D7100, Fuji X-T1 and a D800E that I can do comparisons to each camera similar to that undertaken here. Yes, the Fuji does underexpose, but easily sorted in Lightroom on a batch process. Sometimes I dial in EV +0.66 as an earlier poster stated. In terms of detail, certainly at high ISO values I see little difference between my Nikon D7100 and Fuji X-T1 or E-2. With my D800E it’s a different matter, but I would never challenge my Fuji’s to compete with my D800E anyway. In terms of Chroma noise, I have to agree that Fuji have that nailed when compared to Nikon.
It’s horses for courses of courses, but I rarely turn either make into a competition. I love both Nikon and Fuji in equal measure whether Fuji are “cheating” or not.
Richard
Great article Nasim! I’ve wondered about this for a long time, but I’ve put it down to the difference in the X-Trans sensor. Still, I love mine!
Daniel, I love my X-T1 too! Although Fuji cheats a bit with their files, the output is still amazing and the cameras and lenses are superb. Testing out the new 16-55mm f/2.8. Seems like a solid lens in terms of performance! Just big complaint on its stupid hood – takes a lot of force to put it on!
Rather doubt this “revelation” will cost Fujifilm engineers or its many satisfied customers much sleep. Maybe you could share some supporting references of this “pretty known fact among the photography community” and how it detracts from the performance of Fuji products?
CGW, the above article does not show bad performance of Fuji products. I love Fuji and I did not write this article to discredit Fuji or to make it sound like Fuji makes bad products or decisions. Just showing what I have seen so far. If you read the article, you will see that despite the 2/3 of a stop underexposure, Fuji actually shows very impressive performance. So if Fuji did not go that route with underexposing images, their sensor would still compare favorably with other sensors, thanks to the reduced chroma noise.
Hi Nasim, I think the problem is the use of the word ‘Cheat’ to describe Fuji’s decision to use an alternative (yet completely within standards) output definition. I understand the essence of the article is not to discredit Fuji and does support the fact that even after compensating for the lower apparent exposure the files still look better (cleaner) than cameras with similarly sized sensors but using the word ‘cheating’ reads like a sensationalist headline to grab attention, itself not a crime but one that should be made clear fairly early in the article to not be case. It’s also a bit disappointing that you use the same word throughout your comments and responses in this area of the post. I know you own and like Fuji gear but for someone with your standing on this site to continue to support and repeat that unfounded claim only serves to feed those who feel the need to criticise smaller camera manufacturers, who by the way are the only real innovators in this industry at present (this includes all but the two main players).
Each time you see a headline with the question mark, the answer to the question is “No”. That’s what Peter Arnett taught me nearly 30 years ago, and it is still true ;)
You’re missing the point entirely, this isn’t about bashing Fujifilm’s products.
Well, thanks for finally explaining T-Stops. I’m basically self taught and have only seen the term T-stop used in cinematography. I figured it was some variation on F-stops but I never found out what the differences are.
I remember watching a “behind the scenes” video of the making of the movie GAMER which was shot using the digital system Camera RED. The cinematographer had noticed a variance in exposure values between individual cameras so did tests to determine ASA values and rated each camera as such and wrote the values on each camera.
I’ve always felt that Digital Cameras should be considered as the “Film” since like when we were shooting 35mm, we were buying different types of film to get different results. You partly explained something I noticed that I used to get different exposure results with my film camera than I do with my digital (though all the talk of histograms and other tech talk, I am really clueless. Someday the light will go on). My Nikon D800 seems to have a metered exposure value that is at least a +0.7 which I can correct just by decreasing my exposure value by that amount. But when I’m shooting in a dark situation it’s telling me I have nothing even at 6400 when I remember getting results at 800-1600 with film back in the day and even my Canon 40D gave me something (though sometimes with noise). At least I’ve gotten better figuring out how to set the camera in advance for night time or dark room situations. What I do like about the Nikon is that sometimes using Photoshop’s RAW I can pull an acceptable image from a black image as long as it’s within the 5 levels allowed by the RAW filter.
Hi Nasim,
As you point out in your article there are inconsistencies between ‘manufacturer stated ISO’ and ‘measured ISO’ on my many cameras, not just Fuji. (although the amount with Fuji does seem very high).
DxO labs does these measurements when they test the sensors in camera bodies. Unfortunately they do not test the latest Fuji products.
When I was doing my review of the Nikon 1 V3 I did not notice anything when shooting stills and it didn’t occur to me at the time to compare settings for still exposures. When I was doing some video tests I did notice some exposure differences. I always set everything manually with video to try to ensure consistent exposure throughout a scene and to match up my cameras. My V2s and D800 match up well and I never have to correct exposure between the cameras on video but I would have to adjust exposure with the V3. When I checked out the DxO Labs ISO ratings I found that the measured ISO on the V3 was quite a bit different than on the V2. I’m not sure that there’s anything nefarious going on, but things like this tend to have users think they are shooting at a higher ISO then they really are.
Tom
you build your test using raf and lightroom. the trouble is not with sensor but Ith softWares that doesn’t demosaic well raf at high isos. the phenomenon start at 1600 and is progressive as you push. have a try With silkypix that fully integrate the elements for demoisaicing raf and you will see …
it’ common ar this time with third party softwares.
.if you make a f16 rule test : no troubles
.if you shoot pictures in same light at 200,400,800,1600,3200,6400 raf iso and open them in a third party software you will see darkness comming …
.if you open those raf with silky that integrate the all demosaicing raf process, you will see that differences are much more less and results are betters.
.and if you have a look at those raf on your camera they all look same 2…
it’s more a trouble with software integration than with camera …
waiting for lr 6 to see how raf demoisaicing will be integrated…
of course there ´s always some differences beetwin cameras, lenses and manufacturers but they are not as important as people show just using a software that is good but doesn’t still demoisaic raf perfectly…
best regards :-)
Although the findings are interesting I don’t really believe they are unusual. I understand we are talking about the Fuji line but with my Nikon D750’s (I have three bodies), two seem to always overexpose by 1/2 to 2/3 stops. One is dead on and that one I leave in the studio/lab. I once noticing this and working through the situation I equate it to color slide and negative film from the past. Every time I would order bricks of roll film, either 35mm or 120/220, testing of the different lot number was a must in order to obtain correct exposure and color rendition. Although Nasim is able to explain this phenomenon, it really is just an alert that it happens and good advice for the more novice photographers here in PL. Many of you who look at things like test results on a regular basis when acquiring equipment may have been aware of this and know how to compensate. I think Nasim does a very good job here in explaining what he found and how to make adjustments for it. I know with the two D750’s I use in the field I understand how to adjust depending upon the photographic situation I am in. So would most of you.
Well Canon and others are doing the dual pixel or pixel shift thing to “increase” resolution. Fact is that unless you have a very good sensor, you cannot disguise the shortcomings and shortcuts.
Fuji reports the amount of midpoint compensation necessary in the Makernote tag 0x9650. A raw converter needs to take this into account when forming the “shoulder” on the characteristic curve, but Fujifilm as far as I know never documented this tag and we needed to hack it.
Here are the values for XT-1:
ISO 200 / 0.72EV
ISO 400 / 0.72EV
ISO 800 / 0.72EV
ISO 1600 / 0.72EV
ISO 3200 / 1.38EV
ISO 6400 / 2.38EV
The reason for -0.72 is to give more room for highlights.
Iliah, is this the reason why images that I import into RawDigger look the same on Fuji X-T1 as on other cameras? I loaded the same files, with the 2 second exposure into Lightroom and brightness looked completely different, while it looked the same in RawDigger. I created TIFF files from those RAW files and they again looked pretty much identical!
This is great info Iliah, thank you so much for pointing it out. Now I wonder why Adobe is not reading this tag properly…but I think you and I know the answer to that question :)
Dear Nasim, you may want to check Preferences -> Display options -> Automatic exposure correction for RGB render. If it is checked, we compensate for this discrepancy. As to Adobe, I e-mailed them as soon as the tag was found. maybe they will decide to act. Maybe not.
Iliah, indeed, it was checked. I should have emailed you last night to find out exactly why I was seeing the same images, but I assumed that you were silently compensating for the difference, which turned out to be true. Just didn’t know that it was Fuji’s original intent to do so…I wonder why Fuji did not let everyone know about these tags beforehand!
> Fuji did not let everyone know about these tags beforehand!
Exactly. If Fujifilm are to be blamed for something, it is for being as secretive as all other manufacturers about the metadata in the files. As I repeated many times, the problem is not in disclosing the raw data format, it is not that hard to hack. Metadata is the main problem. It must be openly documented first and foremost.
Iliah, and we discussed this over the phone a while ago…such a big PITA that manufacturers cannot get in line with developers – it is for their own good! When will we see a day with detailed compilation of ALL metadata stored in RAW files? It would make coding so much easier. Programmers would spend time coding, instead of reverse engineering.
There are 2 immediate things – first, all ISO ratings in the standard are essentially not relevant to raw, and so are the ISO settings on the cameras. Second, if one is using mainstream converters, the colour transforms are calibrated the same way the in-camera meter is calibrated; that is, until the midtone is not brought back to the placement as it is in the out-of-camera JPEG, colours are not as manufacturer designed them. The third, not-so-obvious consequence, is that because of the strong compression in the shoulder brighter colours (above L*≈76) are rendered without any accuracy.
Adobe is a behemoth. Wish I could go to their roadshow here in Singapore and tell their “evangelists” that not everything is hunky dory…
Iliah, could you please take a look at the above article summary? I made changes to it and specified your table, if you don’t mind. Thank you so much for your feedback and for letting us know the real reason for this. So many reviewers, including DPR use Adobe for conversions! Looks like DPR needs to update their comparison tool with the above changes to make the comparisons more valid.
“Fuji underexposes its images by around 2/3 of a stop to a full stop when compared to other cameras” – maybe to mention it is about raw?
“we now know the reason why Fuji RAW files appear brighter” – a small typo?
Iliah, that’s what happens when I am a walking zombie after a few hours of sleep :)
I changed the first sentence and also added RAW converters – hope it sounds better now.
Zombies we are ;)
So, adding up the numbers and extrapolating the trend, the sensor is keeping up linearly up to about 2000ISO and then it hits the wall, everything after that point being gain? Correct?
At 2000 it switches from analogue amplification in the camera to digital multiplication in converter.
Is this value the same for other X-Trans camera? For example the X-E2.
The X-Trans sensor’s great at pulling out shadow detail while keeping chroma noise to a minimum, so maybe I’m being generous to Fujifilm, but to me it seems intended to avoid blowing out highlights which are much harder to recover.
David, I think the X-Trans sensor can do very well with both highlights and the shadows…Judging by what I see when normalizing ISO to other cameras, I think Fuji would have done well either way. Chroma noise is indeed nicely taken care of.
The ISO standard for digital camera speed rating (ISO 12232:2006) gives manufactures a choice of five different methods for determining the rating. I’m not surprised that we see quite large differences.
For camera RAW, the ISO speed rating by itself is meaningless/arbitrary. Consider two different cameras using the same sensor: camera A is designed to have 1 stop of headroom above its JPEG clipping level; camera B is designed to have 3 stops of headroom. For a given scene, camera A must be given two stops more exposure than camera B. To ensure this occurs using the inbuilt or external light meters, the ISO speed setting on A must be one quarter that set on B. In the deep shadow areas of the scene, A will have two stops lower noise than B.
So, which camera is lying about its ISO values? Once put into context, neither camera is lying or cheating. What I’ve described is the basics of ISO-less sensors.
Whenever the sensor has a wider dynamic range than that of the scene being captured, we have a choice of ISO values to use. If we need a fast shutter speed then, in effect, we must expose to the left and suffer increased noise. For motionless scenes, we expose to the right by dialling in a lower ISO value to inform the exposure meter so that it can help us to perform this task.
I’ve not yet seen a camera that will display RGB histograms based on its RAW clipping levels; only on its JPEG clipping levels, which is completely useless for recording and processing RAW.
Recording JPEGs is similar to using transparency film (quick onset clipping of highlights), while recording RAW is similar to using negative film (much greater highlight headroom). The ISO speed rating was, and still is, nothing but an awkward compromise to encompass the two very different limitations of JPEGs and transparencies versus RAW and negatives.
If that wasn’t bad enough, we have sensors of differing technologies, number of megapixels, maximum electron counts, and photon efficiencies. This renders ISO speed values even more meaningless for comparing the RAW capabilities of various cameras.
Fujifilm are using Standard Output Sensitivity definition, plain and simple.
en.wikipedia.org/wiki/…_.28SOS.29
Far from plain and simple, in my opinion, especially in terms of how it relates to the RAW clipping levels I was talking about. RAW clipping levels relate to the maximum achievable performance from the sensor, not its ability to perform at a chosen ISO value.
I payed for the original document and do not need to read an abridged copy in wiki.
First, Fujifilm indeed are using Standard Output Sensitivity, plain and simple, and mark it so right in their EXIF. Second, raw clipping levels are only remotely related to “maximum achievable performance”, as those levels depend on the gain applied, different for different ISO settings, while sensor performance is specified for the amount of light, not the gain.
There is no need in making the subject more complicated than it is by trying to involve different methods of stating ISO when the company uses only one.
How does your reply relate to Nasim’s article or to my first comment?
I clearly stated that ISO 12232:2006 gives manufactures a choice of five different methods for determining their ISO rating. Fuji has selected the Standard Output Sensitivity whereas many other manufacturers have not — this is what makes the subject complicated. If all manufacturers used the same one out of the five methods then we wouldn’t be having this discussion.
If you have a specific question to ask then I’ll try my best to answer it.
Directly. 5 different methods are irrelevant here, as 1 is in use.
> Fuji has selected the Standard Output Sensitivity whereas many other manufacturers have not
What are those “many other manufacturers”?
If you understand what is in fact SOS – it is a standard for JPEGs. So different manufacturers are free to interpret SOS as it suits them. All of them using SOS does not mean raw level for midpoint is the same.
Not having been employed by any of the manufacturers, I’ll have to take an educated guess: It is the many manufacturers who’s ISO ratings make it look as though Fuji is cheating with the ISO ratings of its sensors. However, it is not beyond the bounds of epistemic logic that Fuji is correct and that most other manufacturers are cheating with their ISO values. I refuse to make an appeal to popularity or an appeal to general consensus, however, an appeal to expertise is definitely not a logical fallacy.
I consider Nasim to have a high-level of expertise in many aspects of photography and his articles are in alignment with those of other experts, including many of the experts who design cameras and those who design lenses. Kindly note that an appeal to expertise is NOT a fallacious appeal to authority.
What exactly is it that you continually fail to understand in my comments? I’m not saying that Fuji _is_ cheating, I’m simply offering scientifically justifiable reasons for the discrepancy in the ISO ratings of its sensors. Would you rather I provided science- and evidence-based reasons to support the notion that Fuji is deliberately cheating?
Pete, Iliah was simply pointing out the facts relative to Fuji, so please do not get offended :)
Nasim, I’m not in the least bit offended. I’d likely be offended if I relied only on my opinions and experiences rather than on science and data :-)
Many thanks for your excellent and informative article.
Pete, please see the updated summary to the article. Turns out there is no cheating involved – Fuji simply failed to provide its “tag” data in RAW files to developers. Iliah Borg and his team discovered the tag and you can see the results in his comment below, or in the summary section above. Exciting stuff! Makes me wonder why manufacturers do not want to work closely with RAW developers to make their files look as originally intended…we would not have all the issues we have today with converters.
Nasim, my original comment was attempting to clearly demonstrate that no cheating was involved. Perhaps I should’ve worded it more forcefully rather than using my utmost respect for you and your wonderful website.
I consider being allowed to comment on your articles to be a great privilege, not a human right.
Pete, in that case we are all in agreement and you and Iliah taught me something today, which is what this website is all about. Thank you for sharing your knowledge, I really appreciate it!
And please, feel free to write to me any time and your contributions to the forums and comments are greatly appreciated.
Have a wonderful rest of the week and I am looking forward to learning more!
Nasim, my sincerest thanks to you and Iliah because you have both taught me very valuable lessons.
When I ‘talk technical’ I tend to be terse, exacting, pedantic, and confrontational because those were just a few of the essentials required during my career in engineering :-)
“Cheating” they not. ISO ratings are defined for JPEGs, mostly sRGB JPEGs on top of that. For raw converters they have a tag that specifies the compensation. The table is in one of my posts below. If Fujifilm feel comfortable with the noise in shadows caused by overstating ISO and thus decreasing the exposure (real exposure, which is the amount of light) it is the task of a raw converter to render the correct brightness of the final image. As for us, we implemented the reporting of that tag in LibRaw for everybody to use. So did Mr.Harvey in exiftool. Both softwares are free and opensource.
Iliah Borg, I shall leave it to the readers to decide your modus operandi in your comments. If you think that you can teach me something about the technicalities of JPEG; RAW and its converters; sensor technologies and formats; EXIF data; and most of all, photography, then you are hopelessly mistaken.
In my many decades of photography I’ve always admired Fuji products — their awesome films, printing papers, cameras, and lenses. I’m not offended by your insults to my intelligence, however, I feel strongly that Fujifilm Holdings Corporation ought to be offended by your pathetic attempts to support the company. For goodness sake, my comments do a much better job of supporting the company than have yours!
Pete, I don’t think Iliah was trying to offend you at all – he was just stating the facts and perhaps it seemed a bit condescending. Knowing Iliah, he is not that kind of a person. He probably has not had much sleep like myself and he was being short and a bit abrupt. No need to get offended my friend, neither myself, nor Iliah would question your intelligence. We are just trying to have a friendly discussion and perhaps it got a bit too technical, at least for me :)
The assumption that my intention was to somehow support Fujifilm is totally wrong. If anything, I would be trying to support Nikon, Olympus, Sony, Leica, or Hasselblad – as those are my bread-and-butter cameras :) But I do not do even that, not sharing fanboy mentality in the least.
The following is for the benefit of curious readers.
There is no such thing in sensor documentation as “ISO ratings of sensors”. Sensors are not rated in ISO speed units. Often even the word sensitivity is not used, they use responsivity instead (because sensitivity refers to minimum magnitude of the signal that results in specified output) – like 1.4 V/lux-sec at 550nm.
Sensors and cameras are different objects, and using terminology established for cameras while speaking of sensors is nearly always wrong. Same, JPEGs and raw are different objects, with different properties, different metrics, and different testing methods.
ISO speed was never defined for latent images; and raw is a latent image (not a negative, obviously). There is however ways to define ISO speed for raw, like saturation-based; this allows to perform cross-comparisons between different brands and models. But again, such ISO ratings are not about sensors, not about cameras.
There is no such thing as “ISO-less” sensors, too. Near ISO-less, yes, but one needs to be comfortable with that “near” and to know the limits.
With both film and digital ISO speed is defined for processed / developed image. Neither latent image, nor raw image are processed or developed. Change the processing, and ISO speed changes. Manufacturers account for non-standard processing and make an effort to provide guidelines. For example, Kodak specifies ISO 50 rating for T-Max ISO 100 film when processed in Microdol in a certain way; and ISO 800 for the same film when push-processed in XTOL developer. When we deal with a non-standard converter / developer (like Adobe for Fujifilm), it may cause ISO speed different from the rated; and again, like it is with film, one can adjust conversion / development parameters in a non-standard converter / developer to get the rated ISO speed.
ISO speed is a very important parameter, far from being meaningless or arbitrary, because from the practical standpoint it affects noise. Within the range of analogue ISO settings, the higher is ISO, the lower is the noise; contrary to popular belief, where high ISO noise is a misnomer for low-light, underexposure-caused noise. ISO setting also affects colour accuracy, especially for the shadows, like I demonstrated here www.dpreview.com/forum…t/55577930
Thanks for your input Iliah, it’s a pleasure to read and learn from someone who clearly has such a grasp of his subject. I have nothing more to say.
All the manufacturers “cook” their raw data. All the sensor data goes through a lot of manipulation before it ends up the “raw” data in our raw files. Another thought: all sensors have one sensitivity. All the ISO “stops” are either achieved by adding gain to the collected photons or just a mathematical formula.
It really doesn’t bother me if Fuji trends darker. You can see exactly what you’re getting and can adjust to get what you need at the time of capture. It is also worth mentioning that your choice of film simulation will have a big impact on perceived brightness/darkness.
NWCS, agreed on both counts. And it does not bother me either – I wanted to show this difference in the article, that’s all. I also showed that if Fuji did not do this, they would not be losing out big time anyway.
I’m a newbie to these more advanced aspects of photography and got a bit confused about the subject of exposure after reading this article. I thought that, when it comes to “light” in the picture, there is no real difference between changing exposure with ISO, aperture or shutter speed. But it seems that increasing exposure with software is something else? Nasim concludes in a comment here that the fuji exposes correctly and that you shouldn’t increase exposure in the camera but rather with software. But still the original fuji-picture in the example is darker than the sony. Why can’t you than say that the fuji-picture is underexposed and that you should for example lower the shutter speed? I’m a bit perplexed.
Luddis, please don’t pay attention to this article and stick to what you know, since it only might confuse you. The Fuji cameras do not underexpose, so no need to think about the exposure. They only underexpose relative to other cameras – that’s what the article is about.
And yes, you can change exposure in post-processing as well, that’s what the Exposure slider is for. But my recommendation is to do it right in the camera.
You are probably right that I should forget about this article since it appears above my technical sphere. But anyhow I like to visit your site to get inspired and try to develop my photographic technique and know-how. Thanks for your great and diverse site.
It’s still good to read articles even if you don’t completely understand them. Bound to learn something! And someday you might have an “ah-ha! moment” where you think back and say hey I get it now. :)
The “unbiased” ISO rating for raw is saturation-based, recording the mid-tone 3 EV (stops) lower than clipping occurs. Since this one is not used now by the camera manufacturers all the burden to compensate when shooting raw is on the photographer, and it can be done either while shooting or in conversion.
And who gives a crap? seriously guys, was it a slow news day? This is done across he ENTIRE camera industry. I don’t care what they do as long as the Fuji cameras keep kicking butt!
Jorge, many do give a crap to understand the true performance. And if you actually read the article, you would see that I did not write the above article to say negative things about Fuji. Even with adjustments, the Fuji X-T1 looked pretty impressive…
Hi Nasim
Great review, this is an eye-opening review.
Would setting exposure compensation to -2/3 stop on any APS-C camera such as the D7100 would bring similar results as the Fuji shows ?
Luc, exposure compensation would end up changing the shutter speed, so it would affect the overall exposure. And yes, it would look similar, but the settings would not be comparable anymore, just like they were not at the start.
Nice article, going together with my experience. Especially interesting wrt. the tags discovered by I. Borg. One other note related to the appearance of noise. There are many science paper, I read out of fun to understand the random CFA of Xtrans. Mostly they result in s.th. like:
“Our new CFAs yield demosaicked images where the artifacts appear as incoherent noise, which is less visually disturbing than the moir´e structures that appear with periodic CFAs.” Source: Random Patterns For Color Filter Arrays With Good Spectral Properties, Laurent Condat. There are many more papers investigating this, but the incoherence is one of the key issues in my opinion.
Interesting read, I thought it could be attributed partly to the custom color calibration profile that Fuji implements, as I’ve seen the shadows in astia to fall off much faster than Adobe standard, but I’m assuming you tested the cameras in a common color calibration profile
The main reason why I never started buying Fuji is the fact that all their images looked not as sharp as the rest of the peeps out there…. they looked over processed with noise reduction… I looked at the Dpreview charts are the images are just not sharp and they do look the same from ISO 200 to800…. anyway I like Fuji but I will wait…
Forget all that pixel peeping. And in particular, on the DP Review charts. There’s a lot of ‘post’ influence on it as well, caused by their Adobe preference. There are Fuji flaws (see my post) but sharpness has never been an issue for me. Even in the latest LR versions – with particular settings – I get the demosaicing very close to what any other convertor would do. But out of the box, LR is still not where it should be. And that’s not only true for Fuji, unfortunately.
Fuji is pretty darn sharp. You just need a raw convert that supports it- or just use JPEGS in their latest cameras which is surprisingly biting sharp.
Yes and I think they use some noise reduction on RAW files.
As always interesting, Nasim. I must admit as a Nikon F(X) addict and first wave Fuji X user (X100, X-Pro1… X-T1), I have to deal with a love and hate relationship when it comes to the X-firmware, standard X-settings and even ergonomics. To my taste & perception, Nikon is a mostly a bit overexposing and Fuji X is indeed moving in the contrary sense. But there’s more. In critical high contrast situations (take f.i. a street, one side in the brightest sunlight, the other side shadow), Nikon deals a lot better with this kind of exposure/DR challenge than Fuji ever does and it’s not only the full frame-factor. I often found the X-Pro1 to be a fair bit off-road (mostly way too dark, underexposed), the X-T1 is better in this respect but Fuji… but still remains very different, a different ‘black knee’ than Nikon has. Also particular colored LED and fluorescent lighting can force the Fuji-algorithms is extremely strange exposures and white balance deviations, even seems to confuse the AF. Despite the pains I can still live with it as I can accept other camera systems as well, every tool has its pros and cons – Fuji’s biggest merit is the excellent fast glass (while most – even very expensive – DSLR glass doesn’t deal so well with full open apertures and CA).
I almost always compensate -1/3 or even -2/3 on my Nikons and the opposite +1/3 or +2/3 on my Fuji X100S. Thank you for letting me know I am not crazy. :)
Sounds like you are using the histogram and your brain. An excellent combination.
Fuji’s do not underexpose, the midpoint of their RAFs might be set at a lower ‘brightness’ level than other RAW files but once this is compensated in post you’ll find the highlights and shadows are well exposed. If you are always compensating +2/3 you may be definitely almost always blowing out highlights. As Nasim mentioned somewhere on this thread “dialing +2/3 of exposure compensation won’t do anything, because you are just going to end up blowing out the highlights. The Fuji X-T1 exposes properly, that’s not the issue – the issue is when you compare the same exposure across different cameras.”
Sorry, I made that confusing since I wasn’t commenting on the premise of the article. I was replying to EJPB’s comment since his real world experience with the cameras is similar to mine. I’m always compensating on the fly for the exposure that the metering system of the camera is giving me since it is never perfect… no matter what camera I use.
In my experience, much of the time my Fuji X100S seems underexposed and most of my Nikons seem overexposed — when I am actually taking the picture, not in Lightroom. This is based on gut feeling and experience as well as from the histogram, the preview on the camera, and how they look in Lightroom on my computer afterward. I have no trouble with blowing the pictures out or editing them in Lightroom, since after I compensate in the field they look great in Lightroom and just need a little seasoning to taste rather than a complete re-cooking to fix the exposure.
Just as the article shows, on my Fuji files I almost always have to additionally bump up the exposure to compensate for Lightroom not reading the RAF properly, regardless of how I compensated for the exposure beforehand. This adjustment gives me the exposure I experienced in the field, so I totally agree with Nasim and others about this phenomenon.
Well you could be very right about this finding. And my feeling about Nikon very similar to Joshua -1/3 to occasionally -1EV but I try to keep this kind of change in post.
@MegaZ:disqus Since DxOMark can currently not test X-Trans sensors, there’s no way to get a numerical comparison with the competition. But how far away from FFs would you empirically rate it? Thanks
@Nasim Mansurov:disqus Since DxOMark can currently not test X-Trans sensors, there’s no way to get a numerical comparison with the competition. But how far away from FFs would you empirically rate it? Thanks
There is a simple procedure to find out how the exposure meter in the camera is calibrated in regards to raw levels. At each ISO setting you need 2 shots, one containing a blown-out area (like a shiny metal ball under sun, or under a bulb), and the other shot is of some uniform surface (grey card or a white wall) in a spot-meter mode. Bring those two shots into RawDigger (free trial is enough), see what the maximum is on the first shot, and see what is the average level on the second. Now, dividing the second number by the first and taking log base 2 you will have the calibration point (some call it headroom in highlights). Normally it is 3 EV (12.7%). This way you can compare cameras, and use DxO data as the reference for the cameras you do not have.
I don’t understand why this would work out better for Fuji. An image underexposed by 1 stop and then pushed 1 stop in post should look worse than a properly exposed image, should it not?
In general photography that would be true, but when evaluating sensor performance it’s not that simple. It may make for an interesting technical discussion but does it really matter that much in photographic practice?
Interesting information, but, for me, one fact overrides all this measure sting! After a 45 year career as a pro shooter, my Fuji X camera images are the best I’ve shot in my life, good enough for me! After all, isn’t that the point of wanting great performing gear?
In raw optimal exposure is about placing the important part of the scene within the dynamic range of the camera (varies with ISO setting). Underexposing by 1 stop means 1 stop of important shadows in the scene are unusable while the last stop in highlights is wasted. Is that the case?
Mixing up exposure and brightness is a dangerous misconception.
Who want a camera with noice a medium luminance noice reduction always on, even in RAW? (Less noice and less detail.) Thats simply what the x-trans stuff is all about.
Dear Olaf,
What is the source for your conclusion, did you make any FFT analysis?
What I think Fuji might be doing is not actually reducing the sensitivity of the sensor, but subtracting a dark exposure from the raw directly after the shot. This will reduce the overall intensity, but is not really “cheating” in the sense of using wrong ISO values. Just increasing exposure will still lead to overexposure, even if the data recorded is far from the possible maximum after the subtraction.
In an ideal case this will leave nothing but the natural photon shot noise or poisson noise in the final raw, making them look a lot cleaner, especially for high ISO values.
Hold on-the a7 ii is full frame and the fuji is aps-c. The apertures are not equilavent for full frame and aps-c. You lose a stop of light with aps-c!
This meme needs to die, once and for all. An f/2.8 lens transmits the same amount of light to the sensor no matter what size the sensor is. Any differences in light gathering potential come down to the sensitivity of the photosites on said sensor.
Sorry but its also on well established photography sites that say this such as photozone and dpreview. Google “dpreview equivalence” and read the article. Its also quite clear from a few basic experiments on nikon apsc and ff in manual setting. Do you believe a f0.95 4/3 tiny lens really transmits more light than the canon 85mm f1.2 beast on ff!
Sorry Dan, but yes it does, per same area size of any given sensor that is. However, the larger real estate of the larger sensor will result in greater total light gathering by the larger sensor. This translates into cleaner images, but not different exposures. I.e. An f1.8 lens gathers the same light levels regardless of the sensor size, however, the total amount of light gathered by the sensor will differ according to its size. Don’t believe me? Try it for yourself. Rent two cameras, say from nikon, one full frame and one aps-c. Mate them to the same 50mm f1.8. Then set the cameras to the same metering and exposure settings (aperture, shutter speed and ISO). Take a picture with each setup under the same exact lighting conditions. Then check the histogram. Assuming they are cameras of the same generation, you will see the same or similar histogram view. Another example, again be sure to use the same metering and set the cameras to auto ISO in manual mode, both cameras will choose the same ISO (+/- 1/3) for any same aperture and shutter speed combination given the same lighting conditions. I’ve tried it in the past and those were the results.
With respect Mardock, it is you who is, I’m afraid, following a hoary old meme. I thought this old chestnut had been buried some time ago in a previous discussion.
Sorry Mardock and Andres, an f2.8 lens does not transmit the same amount of light to the sensor ‘no matter what size the sensor’. This is crucially flawed thinking.
Just as the effect of the same focal length is not the same on different formats, so the effect of different f-stops is not the same on different formats.
Saying F2.8 on one format is the same as F2.8 on another format is like saying that 35mm focal length on one format is the same as 35mm on another format.
The focal length is of course the same, but the effect of the focal length is different.
It is true that the same f-ratio/f-stop results in the same exposure for the same scene and the same framing no matter what the focal length or format, but in this instance, the framing/format is not the same as we are comparing FF with DX, – so the exposure is not the same. It would only be the same if the format were the same.
You need to remember that the exposure is not the total amount of light falling on the sensor, it is the density of light falling on the sensor or the amount of light per unit area. The exposure is just one part of the equation. The important thing here is the total amount of light falling on the sensor and this is calculated from: Total light = Exposure X Sensor Area. Thus while the exposure might be the same, the sensor areas are different and so the amount of light falling on the sensors is different.
(The sensitivity of the sensor is another story altogether but for the record, the “light gathering potential” of a sensor has nothing to do with it sensitivity. Sensors do not gather light, they just react to light falling on them.)
What fuji is doing is deliberately overstating the ISO as to give the appearance of higher performance per stated ISO. If Fuji renders an ISO of 6400 for the same shutter speed and aperture that register as 3200 for the competition, then Fuji comes out looking like it hits far above its weight class. Another way to look at it is, if for example, Fuji and Sony meter the same scene at 6400, but the Fuji at 1/30 and the Nikon at 1/50 or 1/60 shutter speed (aperture is the same), then the Fuji is getting an unfair 2/3 to 1 stop advantage in light.
When you compare Fuji pictures at the same exposure values (shutter speed and aperture), notwithstanding the stated ISO, the Fuji images look comparable. Make no mistake, Fuji’s take great pictures and the their sensors do a great job with noise. However, they are not miracle workers and they certainly do not compete with full frame sensors when comparing them apples to apples with the competition.
Shame on Fuji and other manufactures who purposely do this. Unfortunately, unless you have a stable of cameras to compare, you are very likely to accept your camera’s stated ISO as fact and will further assume that ISO’s are the same across the board for all cameras.
This brings me to an even greater disappointment. I am as disappointed, if not more, with reviewers. Many, with highly technical reviews and backgrounds, that utterly fail to understand the exposure triangle. FYI reviewers…do not take test shots between different cameras in aperture priority or shutter speed priority! Use manual settings to ensure the manufacturer’s stated ISO does not skew the results between cameras. From experience, canon and nikon tend to state more accurate ISO’. Sony used to through the Nex series. I have found some of their newer models overstating high ISO’s as well (seems to be related to their new approach to pattern metering). Fuji is the worst offender, but Olympus has done its share as well….just to name a few.
Great article particularly with the Iliah update. What I don’t get is why Raw converters then don’t use those Fuji RAW embedded tags to compensate? Something doesn’t make sense.
Question for Nasim (Or Iliah if you want to answer it ;-) ) – for this article I believe you used RAW/LightRoom to compare the Sony and the X-T1 (please correct me here), and the result in LR comes out darker (for not using those compensation tags Iliah pointed out- my guess).
If you shot a JPEG with the camera (or converted the RAW in camera)- does it come out brighter by default? I mean, Fuji certainly would be using their own midpoint tags for their JPEG conversions?
I think that would be a clear data point (without any RawDigger/RAW file tag data investigation) that something is off with the RAW converter (in this case LR).
Both the article and some of the comments were very insightful. Thanks for that!
Hi guys…
I think, that fuji IS NOT cheating – fuji as maybe one from last get it’s job right
fuji says that their ISO 200 is really something about 115-120 ISO – but they dont use it as ISO 100 like others
others cameramakers cheat with its iso settings (maybe except phase one…) – and thats a fact
x-t1 is the my best camera as i ever owned… best iso, manual settings, final results…
of course – thats my opinion
and, i really dont publish my work… i make photos for me, not for others…
sorry for my english
mF
Hello,
Im no expert at all, but i am thinking of the following:
1. If fuji really needed to allow or let the developers to use the midpoint value indicated at the tag, we would see fuji’s jpgs with 0.72EV more brightness than in the ACR developer. So i understand that fuji wants the camera to work with this -0.72EV (either in raw or in jpg). So its correct that we see in e.g. ACR underexposed images in relation with other companies’ cameras.
2. I try to understand why fuji decided on this and im thinking of this: in situations where you do not need to increase the iso in camera, you constantly increase exposure time by 0.72 stops (moving outside of the sensors optimal range), so as to get the correct exposure. Its like exposing to the left and then decreasing the exposure in raw developer, all done by the camera. This way you minimize the noise per photons recorded, trading off the dynamic range (or the highlights) and increase the bit depth/tonality of the shadows. If these are correct, this means that this sensor has great dynamic range, enough to let fuji record an over exposed image and correct it after the capture, to gain some decrease in noise, and tonality in the dark areas. This makes me understand that (if you are for the perfect exposure) its better to expose to the right, since there is enough headroom of the left in the recorded image.
Another trade off of this approach fuji chose is that you have to increase iso speed over 200 more often, when you need shutter speed, which is a minus for me.
On the other hand, when shooting with iso over 200, most often the image has large dark areas, so the previous theory may have even more point.
But in any case all this seems strange to me: capturing an image 0.72 ev beyond sensors optimal exposure, and then underexpose by 0.72 is like saying that fuji doesn’t agree/accept the sensor architecture.
I tested Oly EM10 with 12-40mm and fuji XT10 with 23mm and 35mm f2, and i found that on fuji, with the same exposure settings i had about -0.65 stops difference. This difference is not mainly from lens transmission differences.
Actually x trans sensors works as if its 125 native iso (with “little” highlight headroom and “big” shadow headroom), although its 200 native iso.
What do you think? My logic has any trace of truth or its too complicated to explain to me?
Also, Iliah, how do fuji prevents the highlight clipping with this method ( i would say the exact opposite), since the metering of the cameras, take into consideration this raw exposure handling.
Thank you all for the very interesting comments and the very interesting article!
Sorry to be late to the party. I’m very interested in this. Please can you clarify for me.
Are you saying that the Fuji exposed at say for example, ISO 800 and f8 at 1/125s would give around a stop darker image in RAW than a Nikon at the same settings but that it doesn’t matter because you can boost the Fuji in LR to give as well exposed an image as the Nikon?
Yes but you dont have to boost since the photometer will give you the correct measure so as not to have an underexposed image. So you will have the same photo luminance from both cameras but with fuji you will have more time or higher iso or smaller f number
Sorry if I’m not clear. I know the meter will adjust but it still means going up at least one stop higher in ISO and the results don’t look so good as in a Canon or Nikon DSLR in terms of noise
I take it from this article that Nasim is arguing that the Fuji exposed manually at say for example, ISO 800 and f8 at 1/125s would give around a stop darker image in RAW than a Nikon at the same settings but that it doesn’t matter because you can boost the Fuji in LR to give as well exposed an image as the Nikon. Have I understood his position correctly?
Sorry, also late to this discussion, at least on this site…
There is one aspect in all of these ISO discussions that has baffled me: what is the reason that one should expect that identical numerical ISO values between cameras should yield similar results? Given the fuzzy nature of the ISO concept as it applies to digital cameras, I wonder how realistic that expectation is. Large differences between expectations and reality then often lead to strong reactions, such as the “cheating” accusation, while people with a better understanding and well-adjusted expectations have more measured responses and usually merely note that there are differences in the ISO behavior and sometimes even explain how they can legitimately arise.
We often see comparisons between, say the X-T1 and the Nikon D7000, because they are both based on the same sensor. Then people scream when ISO200 on both cameras doesn’t give similar results in terms of exposure, image brightness and/or image quality.
But ISO200 on the X-T1 is base ISO with minimal amplification. ISO200 on the D7000 is one stop up from base ISO on that camera. The Nikon operates with reduced well capacity and half of its dynamic range at that ISO setting. That isn’t a useful comparison to begin with. Also, because the D7000 has two-fold amplification set, the exposure will of course be lower than the exposure chosen by the X-T1 at the same numerical ISO. Thus the notion that the Fuji “needs” more light to give similar image brightness.
Rather than comparing numerical ISO values, I would suggest comparing ISOs at the same stops above base ISO. So, compare D7000 ISO200 with X-T1 ISO400 (1 stop above base ISO), etc. You will find that the results are now much more similar between the two cameras.
Another useful comparison would be to set both cameras to the same focal length, angle of view, f-ratio, exposure time, and output image size, take shots, adjust them to match image brightness (using either in camera analog/digital amplification or external digital amplification), then assess image quality. This test is grounded in basic photographic (artistic) parameters and should appeal to practicing photographers who don’t want to think about the nitty-gritty details of sensor behavior.
Some speculation as to why Fuji uses digital amplification, even at base ISO:
Amplification is now a mixture of analog and digital amplification. Since analog amplification is lower, chances that the sensor output is clipped are reduced. The subsequent digital amplification of 0.72 stops can be reversed in a raw processor, if the raw processor honors that tag in the first place, thus “magically” recovering up to 0.72 stops of highlights. So, it’s basically a way to help the user preserve highlights.
Also, if Fuji’s ISO scale, based on sensor output, was calibrated to 100 at base ISO, like in other cameras, the subsequent digital amplification of 0.72 stops would shift the scale by ~1 stop to ISO 200 (165 or something, rounded up), which would look remarkably similar to the scale we currently have.
Hi Nixda. I’m sure there might be something in what you say but at the end of the day, I need to crank my Fuji’s ISO up one stop more to get the same result as on my Canon FF and Crop cameras. Most folk buy a camera and read of great noise performance at a particular ISO and expect that to be comparable. People are therefore not really being given very helpful information by Fuji’s marketing department which I am sure is a deliberate strategy on Fuji’s part. It’s hardly surprising they then see Fuji as “cheating”.
I notice you don’t give any results of your suggested comparison in the 5th paragraph. Are you suggesting that a Fuj XT1i exposed at say 1/125s @ f5.6 with a 52mm lens ISO 3200 adjusted in Lightroom to the same brightness as a Canon 5D3 at 1/125s @ f4 with an 85mm lens also at ISO 3200 with brightness left as it is out of camera gives as good noise performance?
Hi William,
IMO, it is misguided to expect that identical numerical ISO values should give comparable results. I do agree, though, that the notion of a “standard” almost implies the existence of such equivalences. However, that “standard” is very fuzzy and has a lot of variables. Unfortunately, neither camera manufacturers nor reviewers have done a good job in properly educating the consumer.
Regarding the specific comparison I suggested, it was not meant to imply that one would get similar results, but to give a recipe for comparing cameras that might be more useful than comparisons based on numerically identical ISO values.
thats true, I has tested on my nikon d800 and canon 6D using same lenses nikon 50mm 1.2, nikon is more brightness on same ISO setting…
This article and the comments are confirmation that Digital Photography is full of whackos. Did people talk about this during the film days? I don’t know. I tended to avoid other photographers back then except when I was in the lab dropping stuff off (and even then most lab jockeys were not photographers but rather printers and retouchers).
All of you need to take a look in the mirror and ask yourself….is your hobby photography, or arguing about consumer goods? I mean really now….”cheating?” What is this, a card game? The Olympics?
In the film days we would test the film/camera and processing to come up with our own personals systems ISO.
Why not use a Standard like the Sunny 16 Rule to compare sensitivity? I love Fuji’s and was very surprised that for my X-T2 it was more like the Sunny 8 Rule!