The new Fuji X-Pro 2 is the first camera from Fujifilm since the release of the X-Series to feature an all-new image sensor. The new APS-C sized sensor is a 24 megapixel CMOS sensor with Fuji’s X-Trans filter array. The new sensor is also said to improve high ISO shooting when compared to the previous generation of camera, despite the 50% increase in pixels. Now that Fuji has gone to 24 megapixels, I thought I’d compare it to one of my other 24 megapixel cameras: the full-frame A7 II.
There are many ways you could compare cameras, but I specifically wanted to see how the new sensor in the X-Pro 2 stacked up against a relatively modern full-frame camera. Now, the A7 II isn’t quite as good as the A7R II in noise performance, but it is similar to other cameras from Canon and Nikon that share similar resolution. To properly compare the cameras, it was important to first establish a fair baseline for the actual sensitivities of the cameras.
Fujifilm is well-known to use a different ISO scale than other manufacturers, and when asked about the difference, you’ll often hear it’s around a stop different, while others will put it at a half a stop. To test where the actual difference is, I mounted the same lens on both cameras (to eliminate transmission differences), in this case, the Canon FD 50mm f/1.4. Shooting that lens at f/2.8 on both cameras, I evaluated the exposure when both cameras were set to ISO 200, f/2.8, at 1/2 second. I then compared the RAW files in Lightroom, and boosted the Fuji file until the measured values of the white, middle gray and black patches on my color checker matched the Sony file exactly. With these two cameras, the difference is exactly 0.60 stops.
The X-Pro 2 sensor, like many modern sensors, is ISOless. This means that a shot underexposed and pushed in RAW conversion will yield the same noise output as a shot taken at the higher ISO and exposed correctly. To compare apples to apples, I shot both cameras with the exact same exposure settings, starting at 1 second, f/5, ISO 200, and shortening shutter speed by a stop and increasing ISO one stop (1/2s, ISO 400, 1/4s, ISO 800, etc) throughout the ISO range. The Fuji files were then all brought up 0.60 stops in Lightroom to match the brightness of the Sony files, such that the compared images are exposed exactly the same, with the exact same image amplification.
The Sony utilized the Sony Zeiss FE 55mm f/1.8, while I shot the Fuji 35mm f/2 WR on the X-Pro 2. Framing was adjusted to match due to the slight difference in field of view between the two lenses.
I took 100% crops from each image, in two locations: an area of fine detail in the center of the frame (the sketch of the Guggenheim Museum), and the right two rows of my color checker, to check solid color noise in mid-tone and shadow areas. Click the image below to open. Be sure to click to magnify the image so you are viewing it at 100%.
When examining the above crops for noise, if you’re like me, your first reaction is that there isn’t that much of a difference between them. The two cameras look very similar from a noise standpoint at lower ISOs, and they resolve similar amounts of detail, with maybe a very slight edge to the Sony. The noise differences don’t become truly visible until around ISO 3200, where the Sony shows slightly finer grain structure and slightly cleaner detail. Both cameras retain good detail up to ISO 12,800, while the Fuji loses a small amount of detail at ISO 25,600. At ISO 6400 and below, the two cameras are extremely close in noise performance…I’d estimate 1/3 stop difference or less. At ISO 12,800 and 25,600, the difference is a bit more pronounced, but still less than a half a stop difference, which I think shows a great performance from the X-Pro 2.
However, the crops above don’t tell the whole story at the highest ISOs. While the color fidelity of both cameras is fairly impressive as ISOs climb, it is particularly neutral with the X-Pro 2, which shows no noticeable color cast as ISOs increase. But most telling is in looking at the whole images, where the dark background of the images show distinct differences. The X-Pro 2 maintains good color control here and shows no visible banding.
Now take a look at the Sony image at ISO 25,600, below. While the well illuminated areas show slightly less noise, the dark background shows some pattern noise and maybe even some slight banding, with a slight purple color cast. There’s even what appears to be some sensor heat artifacts in one spot along the top edge, to the left of the book, that is showing some purplish blobs. Comparing both full size images, I prefer the overall look from the X-Pro 2 at this ISO.
Looking at the performance as a whole, the A7 II has a very slight edge at medium to high ISOs, but it’s a small lead. This is quite impressive for an APS-C sensor. Well done, Fuji!
Compared to the Fuji X-T1
However, the older Fuji sensor already was fairly close to this level anyway, through most of the range. Fuji claims that the X-Pro 2 has notably improved noise performance when compared to earlier cameras, and some reviewers have said the same. How big is the difference?
I shot the X-T1 at the same time, at the same settings, using the same 0.6 stop push as the X-Pro 2 files. Since the X-T1 won’t record RAW files for ISO 12,800 and 25,600, I underexposed them at ISO 6400 and pushed the files 1 and 2 stops respectively to yield an ISO 12,800 and 25,600 image. To compare on equal resolution, I reduced the images from the X-Pro 2 to match the X-T1’s 16 megapixel resolution, and the results are below. For this, I’m just showing the ISO 3200 shots and above. I saw no appreciable difference at lower ISOs.
Hmmm…There’s a difference, but it sure isn’t large, especially at ISO 3200 and 6400, where it’s barely perceptible. The extra resolution of the X-Pro 2 shows a bit finer detail at these ISOs, even when downsampled, but noise performance is very similar. However, at ISO 12,800 and 25,600, the X-Pro 2 shows it’s stuff. There’s a noticeable detail advantage (again, even downsampled), and the noise is finer grained on the X-Pro 2. Still, it’s maybe a half stop difference at ISO 25,600 and maybe 1/3 stop difference at ISO 12,800. It’s great to see that the X-Pro 2’s added resolution doesn’t make for noisier images, but it’s also not a massive improvement either.
So there you have it. It’s just one test, but I think it shows that the X-Pro 2 packs a very good sensor. I’ll have my full review of the Fuji X-Pro 2 in the coming weeks.
70 thoughts on “Fuji X-Pro 2 vs. Sony A7 II: Noise Comparison”
Jordan, since I have not used Xpro2 or Xpro1 for that matter (own XT1) but have used X100 series cameras, I have a question. On X100 even if you start adding hoods, all of the sudden Optical Viewfinder starts getting blocked. You add the converters and then it is blocked pretty bad and you need to switch to EVF.
Fuji Marketing as I saw in all the Fuji Photographers videos and tests, shipped that XPRO2 to testers with 35 F2 lens. Is it a Marketing trick because that lens is quite small and narrow and wouldn’t block the optical viewfinder?
And then if users start mounting 56mm and longer/larger lenses [I have not seen a single video of anyone with these larger lenses on it] , all of the sudden that optical viewfinder gets blocked and becomes of very little use, so one has to switch to EVF, and then the whole idea of optical viewfinder is questionable, i.e., it only makes sense if you use small lenses on the camera???
The OVF does get blocked by larger lenses, and one advantage of the 35/2 is the small size that doesn’t. I chose the 35/2 for this comparison because it’s very sharp, and I have a very sharp Sony lens for a similar FOV in the 55/1.8. However, while the lens is visible in the frame with lenses like the 23/1.4 and 56/1.2, they’re still usable. The 56 blocks part of the frame, but not the part where the smaller frame lines are.
The OVF is most useful between 23 and 56mm. Longer and shorter focal lengths aren’t ideal, and the EVF is more useful. Personally, while I like the OVF shooting outdoors with moderate focal lengths, in most scenarios, and indoors, I prefer the EVF. Luckily, the EVF in the X-Pro 2 is quite nice.
Thank you for your response. what about 16-55, 18-135 and 90/2? How much of these lenses are blocked in OVF?
All three will intrude. The 90 will work just fine, though because the framelines are a small enough section at 90mm that they predominantly fall outside the image of the lens. The 16-55 and 18-135, though I wouldn’t plan on using with the OVF. Even the 18-55 blocks quite a large portion of the OVF at the wide end (and it still blocks it at the long end, but again, the frame lines are small enough here to be OK framing at 55…though the 16-55 wouldn’t be good for most of the frame. The 18-135, you’d have to manually change the OVF magnification just to get framelines that would work for the whole range.
Here’s a quick shot through the OVF with the 18-55 mounted and set to 18mm…you can see how much it cuts off the wide end. The 16-55 would be considerably worse at 16mm. http://admiringlight.com/2016/xpro2_ovf.jpg
Thank you! I also have the 18-55 as well. This thing at the corner would annoy me greatly. Another reason why XPRO2 is not for me. XT2 on the other hand will be mine, hope I don’t have to wait for it too long.
The beauty of the Xpro2 is it’s versatility. If you’re not a fan of the OVF in the first place than stick to the XT1 and XE2. Personally I feel the Xpro2 is more of a street camera and the lens focal length fall within the traditional Leica M film camera focal lengths. It’s such a nice camera. I own the XT1 and the X100T. Different tools for different jobs…I love them all! 🙂
Instead of a hood, use a step-down ring. I used a 49-28 on my X100’s. No vignetting, no ovf intrusion. On my xf18 and xf35/1.4 I also use step-down rings.
reducing resolution = reducing noise
And your point is? To compare cameras of different resolutions, it makes no sense to compare noise at the pixel level. You take an image and you choose an output size, whether that’s on screen or a print of a certain size. When comparing the X-Pro 2 to the X-T1,a print would result in using more pixels for the same output size…so each pixel makes up a smaller portion of the image (thus making any noise smaller and less noticeable). For web display, you reduce to the same output size…and the higher resolution file will have noise less apparent. It’s why the Sony A7R II shows similar noise in a final image to the A7S II, despite the A7S having better per-pixel noise.
To examine noise of the final image, you need to normalize output size….thus resizing to match output size of the X-T1 (you could also enlarge the X-T1 files to match output size). Looking at 100% noise with cameras of different resolutions would be comparing different crops of the same image, or artificially reducing the sensor size of the higher resolution image.
My point is that. I am not a professional photographer but it seems to me that using this method to compare two sensors it is not a good idea because the results can be misleading. You reduced the images from the X-Pro 2 to match the X-T1’s 16 megapixel resolution and you reduced the amount of details which caused reducing the noise. So to my mind the results are not comparable. In other words. Is it possible to interpolate the image from X-T1 to the same resolution as X-Pro2 have ? And will the results be comparable ?
Downsampling improves the image quality of the higher resolution image, while upsampling decreases the image quality of the lower resolution image. Both are viable methods of making the comparison, but upsampling is general impractical and uncommon, because it has no actual benefit (garbage in / garbage out), whereas downsampling is very common.
Jordan’s methodology here makes complete sense.
Probably – but that is a totally valid approach – since this is what matters at the end.
You should not be interested on the pixel by pixel noise at 100% zoom, but on the image quality you will receive at the end when printing to a specific target-size.
It does not matter, if the end-result is better, cause you had less noise in fewer pixels – or cause you had more resolution to reduce to the output-format. If the end-result is better – your sensor was better for the job.
Both cameras he used has the same pixel. So down-sizing is the same for both.
It’s good to see the edge of entry level FF being smoothed out over years of development.
are those gordys camera straps I see? I love those things, anyway, its nice to see the ISO being pushed a little further in an already great camera system.
Yes! I love my Gordy’s straps.
Very nice comparison
1. Fuji uses the tag “RawExposureBias” to instruct raw converters to brighten images by 0.72 stops (at ISOs 200-1600). As far as I know, LR does not honor this tag. When applied, the images should come out at the same brightness as others. You bumped up the image brightness manually by about 0.6 stops, which matches the 0.72 stops that Fuji specifies. Or did you apply the 0.6 stops on top of the described 0.72 stops?
2. When both cameras are set to ISO200, the Fujis have the full usable well capacity and thus dynamic range available, whereas the Sony applies two-fold amplification and thus has half of its dynamic range available. Something to keep in mind when comparing cameras based on the same sensor that have different base ISOs (e.g., D7000 and X-Pro1), but it’s a bit trickier with sensors based on different technologies.
Seems to me (my eye) that the xpro 2 looks a little flat compaired to the sony?
I think you have a defective Sony A7II. My original A7 doesn’t show any kind of purple cast on high ISO shots, after seeing the same “heating” and purple cast problems on a friend’s new A7II, I recommended he traded for a new one without any issue. I think Sony as a severe quality control issue going on, but like Canikon they might get away without anyone acknowledge it. Regardless, nice article, keep up the good work.
I do not have a defective A7 II. This test is perfectly in line with the ISO test I ran with a review sample A7 II last year. Also, some amp glow is not uncommon at very high ISOs. I’ve put about 15,000 frames through my A7 II. It’s just fine.
Jordan I was not question the test at all my friend. I’m not bashing Sony, it’s just something that crossed me after seeing several A7’s, even at 3200/6400 ISO. Glad your’s works fine.
This might skew a little too off-topic, but I’d like to ask what workflow you used to color balance these two cameras?
I’ve got the same color checker mini, and use it regularly to adjust tones and colors for images from a single camera. But I’ve never attempted to get, as you did, Sony to match Fuji. What’s the trick? Thanks.
A very interesting comparison for noise is to focus on the black colour square by comparing the amount of white pixels present in the high iso range images. The white pixel noise is almost 2 stops better on the X-Pro2 compared to the X-T1 and maybe a stop better than the A7II.
Very nice work doing a test like this.
Do you think that when you say
“Now take a look at the Sony image at ISO 25,600, below. While the well illuminated areas show slightly less noise, the dark background shows some pattern noise and maybe even some slight banding”
could be caused by a more blurred background from FF to APS-C?
Nice article, but as nixed already mentioned, you are not really comparing apples to apples. Fuji is using ISO 200 as their base ISO and Sony is using ISO 100. If you now compare both cameras at ISO 200, Fuji is using full well capacity and full dynamic range, whereas Sony uses only half well capacity and reduced dynamic range.
At ISO 400 Fuji uses half well capacity and Sony uses quarter well capacity and so on. This gives always Fuji more edge in the comparison if you just use same value ISO-setting.
I’m really not sure what you’re talking about. The cameras were set to ISO 200…shots were taken. Then the cameras were set to ISO 400. Shots were taken, then ISO 800, ISO 1600, ISO 3200 and so on. I didn’t push the base ISO files for all the comparisons. Unless you know someone who takes pictures at base ISO all the time and then underexposes by 6 stops when they need to shoot at ISO 6400, it’s a moot point. The only pushing of files that was done was bringing each of the Fuji shots up .6EV on every shot to match the output of the Sony. Otherwise, all images were taken at the same ISO values (as marked in the chart…200, 400, 800, etc), the same shutter speeds (1s, 1/2s, 1/4s, 1/8s and so on as ISOs increased) and the same aperture. It’s hard to get more apples to apples than that.
You claim that “compared images are exposed exactly the same, with the exact same image amplification”.
But that is just simply not true. Fuji has named their base ISO as 200. Sony names their base ISO as 100. Now if you use compare those cameras at the same ISO at any given value, Sony is always using double the signal amplification. That’s what is wrong with your test.
Guess why Fuji is using higher ISO-range compared to Canon, Nikon or Sony? They want to see comparisons exactly like you have done. People think that ISO-values are absolutely true even though they can be anything the manufacturer wants them to be. At some level I wish that other manufactures would issue a firmware updates, where every ISO-value in camera menu is doubled. For example Sony A7 mII would have a ISO range marked from 200 to 51200. Fair play, eh?
And by the way, as much as you like to think no modern camera is yet truly isoless. 14 bit A/D-converters are just not enough for that, even if read noise was negligible at base ISO. So there is still a difference whether you boost up the signal by ISO-change in camera or at post production with the EV-slider. In your reply to me you seem to actually understand this, but in your main article you give impression that you don’t. That’s slightly incoherent.
What does it matter what the base is? The values for exposure are identical, resulting in the same exposure at the same exposure time. In what world would you compare SONY’S ISO 800, 1/8S shot to Fuji at ISO 1600 and 1/16s? How in the world would that be a fair comparison? That seems to be how you would have run the test.
If there’s low light, I’m going to set the camera to a certain ISO to get the shutter speed I need. The cameras were set to the same ISO, exposed at the same shutter speed and the same aperture. What Sony has as their base is irrelevant, as you will still need to do that amplification to get the exposure needed for that lighting condition.
Of course base matters, because it tells you how much analoge amplification is applied at any given ISO. At least in principle, because each manufacturer can use combination of analogue and digital amplification at RAW-level. But it should work in the way that doubling the ISO-value from base will double the analogue amplification. Quadrupling the ISO will quadruple the analogue amplification and so on. And applying the analogue amplification will change the signal to noise ratio and the dynamic range. Digital amplification in post won’t change that, it will only change the relative brightness. Because these cameras have a different base-value, you simply can’t compare them in any apple-to-apple-way if just use the same ISO-setting.
If you really wan’t to compare these cameras at similarly exposed and with the same amount of amplification you have to use same exposure setting but always a half smaller ISO-value in Sony. Then it will be a fairer comparison. It’s still not a scientificly accurate, but still a better way.
So, you’re telling me that a fairer Comparison is to underexpose the Sony shot by a stop? Because we all know in your shooting, you’d underexpose every image by a stop, right? Your argument has absolutely no basis in real world photography. “Well, I’m shooting Sony today, but I regularly shoot Fuji. I guess I better use one stop lower ISO and underexpose everything.”
How in the world would underexposing the Sony shot then bringing it up a stop in post do anything to help? If this truly provided a cleaner image, don’t you think Sony would program the camera to do this anyway? Your argument is simply nonsensical.
Or alternatively forget the ISO-comparisons and use just so called ISO-invariance-test: use just base ISO and take a picture in poor light conditions. Then in post correct the exposure with EV-slider. Do this for both cameras with same exposure settings and then eyeball the noise differences. This test is what I would call a fair comparison.
Because that’s how we all shoot, right?
Pavev made a valid point. ISO-invariant does matter for your pixel-peeping test.
And You should answer based on the technical question raised, not based on whether it’s related to practical photography, because after all this kind of tests have nothing to do with photography either.
i’m an owner of both system and all i can say is a7ii files are much better in terms of noise.try increasing iso of Fuji with 0.6ev,then let’s check the results..dont push files in raw…also, i’ve never seen such color casting on my a7ii with high iso’s, i think you’ve a lemon there.
btw, iso3200 of a7ii is clearly much better than xpro2 iso3200…
I have read all of those ISO-less articles on the web, there is no proof or an example that suggest PURE ISO-less behavior in either one of those sensors. this is becoming a myth on the web. every serious article points out that it is close to ISOless but it is not ISO-invariant in practice, not sony and not fuji.
you started right, stating that a baseline should be established, but your choice to compare exposure compensation of 0.6 after the shot was taken does is not proper, you should have compensated the 0.6 exposure before taking the shots.
If you can honestly prove that they the differences may start to show at such small pushes, you might have a point. Small differences might be seen when pushing 4-5 stops or so…. Not at 0.6EV.
It is not up to me to prove it.
You are claiming those small pushes do not amount to anything, to prove it you would need to measure it, but lets say that we dont have the tools to measure, at the very least you should do the proper experiment where the ISO is compensated before the experiment and show results for before vs after. Atm this is a flaw in your paradigm. obviously there are differences, can we “See” them? we may or may not. can we measure them? I believe there is a way to measure it.
In a proper academic paper, if you are assuming something based on prior knowledge, it has to be solid. I understand this is the internet, but for the experiment’s credibility, you should re-run it if you can.
Also note that I have seen the latest fuji pro2 ISO-variant tests on dpreview, they clearly ignore noise pattern and just look into noise strength, when you dial up the exposure, sharp edges become less pronounced (at least in their examples) and they do not acknowledge it.
Compare to your hearts content. I see no difference. One side is X-Pro 2 at ISO 2500 at 1/100s, the other is ISO 1600 pushed 2/3 stop in post. http://admiringlight.com/2016/xp2_iso.jpg
The Fuji files should be shot up 0.60 stops to match the brightness of the Sony at each ISO.
Taken from your comparison and isolated. look at the edges, they are not kept on the right, and are far better on the left.
This is what i was saying earlier, it hurts edges and no one is acknowledging it, only contributing to the myth of ISO-less sensors.
Jordan, did you delete my reply with that comparison image? (that shows there is a difference)
I never saw a reply. In any case, my test is what it is. Take it or leave it.
here it is again https://c2.staticflickr.com/2/1578/25615165872_00f0dc3d36_b.jpg
the edges are clearly affected, it means that there is a difference in perceived sharpness.
It means nothing of the sort. The 2500/1600 shots posted in my comment were taken handheld since I was just interested in seeing a quick comparison with respect to noise. All you’re seeing here is a slight difference in angle that is exposing more of the white portion of the color checker mat. If it were a sharpness difference you’d see it similarly affected on the top and bottom of the square as well, but you don’t. And the fact you needed to enlarge it so much to see a difference anyway kind of proves my point. You’re grasping as straws here. I won’t be replying again. Take my test for what it is. If it’s valuable to you, great. If it’s not, move on.
both messages say “Your comment is awaiting moderation.”
I’m following up on my earlier comment and on Pavev’s comments:
You ended that exchange with “and that’s how we all shot right?”. Perhaps not, but perhaps we should. People do indeed shoot at a certain ISO, and they expect that the same ISO on a different camera gives similar results. Unfortunately, that is simply not the case. There is no standard that would even imply that that is the case. Manufacturers, and then reviewers, have perpetuated that assumption, and it is now engrained in a lot of people’s mind, but it isn’t justified.
Film shooters indeed stepped up to a scene and asked first “what ISO am I going to use?”. With digital, that is no longer necessary, and it is in fact counterproductive. Photographers should be educated better in that respect; it’s high time. Instead, approach a scene with “what framing, DoF and motion blur do I want?”, then set the camera up accordingly, snap the image and worry about brightness (ISO) later. That would get us back to true fundamental aspects of photography, not something about sensor technologies. With the same angle of view, DoF and motion blur as well as, ideally, print size, one can then really compare images for noise, SNR, etc. and make meaningful comparisons that matter in practical photography.
In any case, it doesn’t really matter how one arrives at a final brightness, the amount of amplification, whether done in camera and/or externally, determines what ISO one has (as long as one is looking at JPEGs, because ISO is only usable in the context of JPEGs). Thus, when you increase the amplification of the Fuji images by 0.6 stops, you re effectively increasing the ISO by 0.6 stops. ISO400 becomes ISO600 and so on, and you’re no longer comparing the correct ISO values.
But it doesn’t really matter, provided you indeed use the same amount if light (in the case of the same-sized sensors) or equivalent amounts of light (in case the sensors are different). There is really no need to attach any pseudo-quantitative ISO numbers to images, because that causes more confusion than they help.
The reason Fuji needs to be shot (or adjusted) by .6 stops is because it exposes that far lower at the same marked ISO. it would provide a marked advantage to the Fuji to shoot at marked ISO vs marked ISO because Fuji’s ISOs have an actual sensitivity with regards to the final exposure of 0.6 stops less than what Sony marks their ISOs.
I’m not sure I understand. I thought you are dialing in the exposure (amount of light falling onto the sensor) manually by setting aperture and shutter speed. If you have equivalent exposures between the cameras and you brighten the image of the Fujis in post, you are still increasing the effective ISO, and what’s printed on the dial isn’t the correct ISO anymore. ISO corresponds to the sum of all amplification steps performed, not just the camera’s but also what the user is dong in post. That is another reason why comparisons of numerical ISO values (as printed on some camera dial or shown in some menu) are tricky, because brightening by the user often isn’t taken into account. Finally, when shooting raw, ISO doesn’t exist. The ISO value that is set in the camera then simply represents the amount of gain applied during sensor readout. That amount plus any amount of digital amplification done by the JPEG engine, by virtue of applying tone curves and/or doing brightness adjustments, will then make up the final ISO for the image.
I set the cameras to the same ISO, shutter speed and aperture. I raised the Fuji files in post because Fuji underrates their ISO values. So he’s, it would be the equivalent of a higher marked ISO on the Fuji files, but it’s the same amount of light per area hitting the sensor, and the ISOs matching to produce the same exposure.
So, you just need to increase the ISO values in the captions to the images by 0.6 stops. Nominal ISO400 becomes actual ISO600, etc. Or if you want to retain the nominal values, you could perhaps add a comment to that effect. Most people don’t realize that adjusting image brightness in post changes the effective ISO.
@nixed – I’m not sure you’re interpreting this issue correctly. Fuji cameras underrate their ISO relative to the other cameras on the market. It’s not so much a matter of nominal / indicated ISO 400 becomes actual ISO 600, as nominal / indicated ISO 400 becomes actual ISO 400, because the actual indicated ISO is only ISO 300 at the time the camera makes its exposure. Ordinarily, the Fuji camera’s meter will compensate and leave the shutter open slightly longer than other cameras when set to the same indicated ISO, but because Jordan had them in manual, you just end up with a slightly darker exposure with the Fuji, that is then exposure matched in post to provide the same amplification or “real ISO.” The indicated ISO on the camera is ultimately meaningless unless it’s a true standard that lets you compare between other cameras.
In any case, once again I am in complete agreement with Jordan on this one. Kudos to you for your sensible methodology. It’s frustrating that you need to do post-production exposure boost in the first place, but it makes the comparison much more valid, and it’s the extra thoughtful steps like this that you take that keep readers coming back to your blog for more good information.
@Andrew: totally agreed on “The indicated ISO on the camera is ultimately meaningless”. However a lot of reviewers/bloggers use exactly that number to compare images.
But I still believe that total amplification, including that added/subtracted in post, needs to be considered.
Imagine one does a comparison between analog and digital amplification: first, one would shoot an image at, say, ISO1600 with a certain aperture and shutter speed; second, use the same aperture and shutter speed but shoot at base ISO200, then amplify the data digitally by 3 stops. Would you call the latter an ISO200 shot, just because the dial was set to 200?
I think the easiest way around this issue is by showing the total amount of amplification, and if one is really thorough, specifying the amount of analog and digital amplification in the camera as well as digital amplification in post. So for the above example, both final images have the same amount of amplification (3 stops).
A complication for all of this is what is done in post processing in addition to simple brightening, such as applying tone curves, because they affect the brightness, and it’s not always easily quantifiable how.
“I think the easiest way around this issue is by showing the total amount of amplification, and if one is really thorough, specifying the amount of analog and digital amplification in the camera as well as digital amplification in post. So for the above example, both final images have the same amount of amplification (3 stops).”
This is effectively exactly what Jordan has done. He has used the same aperture and the same shutter speed so that both sensors are exposed to the same amount of light, and then matched the brightness in post so that both are showing the same amplication. In this way you’re truly comparing apples to apples, rather than apples to oranges when Fuji chooses to use an underrated indicated ISO scale.
Obviously that neglects differences in lens transmission, but you can’t take everything into account, and those are practical differences that will be relevant in real world shooting anyway.
And in general I agree with you, an ISO 200 with +3EV in post is for all intents and purposes an ISO 1600 shot with slightly different characteristics, depending on the ISO invariance of the sensor and ADC…
Thank you for this great comparison, Jordan! It was a great read and in my opinion this exactly demonstrates the results that someone like me would get in everyday usage. With these ISO-less cameras I mostly just choose aperture and shutter speed (with auto-ISO) and tweak the exposure later on in the RAW conversion. That would yield precisely the results that you show here. Again, thanks!
One more thanks for the comparison, Jordan. I shoot both, Sony and Fuji X, have done my own comparative shootouts in the past and I’m d’accord with your findings and your approach from a photographers standpoint.
I only hope the critical voices don’t hold you off from writing further articles like this one. Keep it up!
Instead of a hood, use a step-down ring. I used a 49-28 on my X100’s. No vignetting, no ovf intrusion. On my xf18 and xf35/1.4 I also use step-down rings.
Jordan, thanks very much for very informative comparison! I use Fuji and have been considering going for a Sony A77!
Would it be possible to add an m43 camera into the mix and add another line of comparison crops? Something like Olympus E-M1?
Thank you again!
I am SO pleased that you took the time and effort to do these tests, especially the comparison between the X-Pro2 and XT-1, as I just decided to acquire the X-Pro2 as a complement to my XT-1. (I was originally interested in the upcoming XT-2, but frankly, was too intrigued by the new sensor plus all the features of the X-Pro2 to wait! I hope I didn’t make a mistake, as it’s the first “Rangefinder” style camera I’ve ever used, and I shoot landscapes . small products and macro; not people or street.) Though it’s nice to see that the X-Pro2 image quality can “hang” with the FF Sony, the XT-1 comparison was more relevant to me, as I’ll be swapping the same lenses between the two bodies, and it’s nice knowing that color and other image characteristics aren’t wild different. Of course, being a “sharpness fanatic”, I will appreciate the extra detail in the X-Pro2 images! As for the high ISO comparisons, it’s only interesting to me on an academic level, as I try to always shoot at ISO 200 when possible.With respect to the more “critical” comments, I’m not an engineer or physicist, so I can’t pretend to fully understand their arguments. However, I am a photographer with about 35 years’ experience behind a camera, using formats from 35mm to 8×10, so I can safely say that I recognize Image Quality when I see it. Thus, I totally agree with your logic behind your testing, as I’m more interested in “real world” results than whether a “pixel well” is half-full or quarter-full”. Thanks again for your always inspiring reviews and articles!
Well, you can argue all you want but the fact is this test is wrong because Fuji lies with their ISO and this is WELL known but since people like you still do articles like this, then its working for them because people reading this may actually believe that this Fuji APS-C can actually match a FF in noise performance wheen that is NOT true at all.
Well, did you actually read the article to see that I directly measured the ISO and exposed at the exact same shutter speed and aperture completely leveling the playing field?
Since we know there’s a 0.6EV diff, instead of shooting both at identical settings and pushing the Fuji file after what would happen if Sony uses a lower ISO value and push 0.6 stops in post instead?
Sorry I meant to say 0.4.
So same aperture and shutter speed, Sony with ISO 100 +0.4 in post vs Fuji at ISO 200.
I wonder if there will be in difference?
Since I’m thinking to change my a6000, I’m not sure to change the system and go Fuji route (XT2) or stay with Sony and get an A7II. My concern is about how the fuji sensor manage a low light situation. Since I have also a X100T, my main concern with this camera is that it doesn’t handle very well low light situations, but I love the “old” approach of dials and manual controls. I saw that on the new a6500 there are two more buttons but no additional dial (rear and front dial, and an aperture ring, like on XT2 are, for me, things that speed up the process of – enjoying – shooting photos), that’s a bummer for my kind of shooting.
So, what’s the difference in low light between these two cameras you take into consideration here at the same ISO, the same aperture and the same shutter speeds?
sorry I didn’t read well, my question was pointless. If I undestand the article, the difference between the two is of 0.6 ev, then the XT2 is 0.6ev much “darker”?
It’s not darker per se, it just overstates the ISO by 0.6EV. So, if you set both to ISO 800, the Fuji will meter to require 1/2 to 2/3 stop slower shutter speed for the exposure. This test equalizes that by exposing darker and pushing in post, but you can also increase ISO by 2/3 stop and get very similar results at the same shutter speed and aperture. If you don’t adjust and shoot at the same shutter speed, aperture and ISO setting, then yes, the Fuji will be darker, but metering isn’t affected, so shooting normally you’ll just experience the need to bump ISO a bit higher than you would on the A7 II to get the same shutter speed.