The Bokeh Effect: How Sensor Size Affects Background Blur
Of all the things that photographers argue about in our secret monthly meetings, sensor size and its impact on our work is perhaps one of the most heated topics that can come up. From the true “bigger is better” snobs (“Sensors? Bah! 8×10 film is where it’s at!”) to the ones who prize portability above all (“Micro Four Thirds rules!”), the debate between advocates of MFT, APS-C, and full frame sensors often reaches religious fervor.
Contentious topics related to sensor size include resolution, high-ISO performance, and dynamic range but the quality and characteristic of bokeh, or out of focus backgrounds, is perhaps the most fiercely debated. While there’s no contest that the bigger sensors can clearly produce much smoother and, well, blurier (not a word, I know), it’s also an unfair statement that the smaller sensors like the ones in Olympus and Panasonic Micro Four Thirds cameras can’t produce good bokeh.
The Prerequisites for Understanding Sensor Sizes (Updated)
Now, before you get into this article, if you have questions about what crop sensors are, how they work, etc., you want to read a few of these articles:
- New DSLR Owners: What You Must Know About Full Frame vs Crop Frame Sensors Before Choosing a Lens
- Understanding Sensor Crop Factors
- Transitioning from Point-and-Shoot to DSLR: Understanding Full Frame vs Crop Frame Sensors
- Best Wide Angle for a Crop Sensor Camera
- Introduction to Full Frame vs Crop Frame Sensors Plus Great Sensor Comparison Resources
- Sensor Size and Filmmaking: Choosing the Right Camera for Your Project
- Aspect Ratios and Sensor Sizes for Video Beginners
- Metabones Canon EF to Sony NEX Speed Booster Adds Full Frame Versatility in the Crop Sensor Video World
- Intro to Depth of Field in Filmmaking
These articles will give you a good understanding of what crop sensors are, and what using a crop sensor camera implies, for the most part. In this article, we’re going to drill down to one specific thing. We will take a look at just how the size of your camera’s sensor affects the bokeh characteristics of your image. To do this, we devised a pretty simple test.
Test Process and Gear for Bokeh and Sensor Size
We took a resolution chart printed on standard matte paper and stuck it on a couch in our San Carlos office’s lobby. In the foreground are a couple of glass terrariums, and our focus point is on the top of one of them. The distance between the subject and the camera is about 3-4 feet, while the distance between the terrariums and the back of the couch is about 4-5 feet. Here’s the equipment list we used.
The list is fairly straightforward. We used three cameras:
- Canon 5D Mark III (full frame sensor)
- Fuji X-E2 (APS-C sensor with 1.5x crop factor)
- Olympus OM-D E-M1 (Micro Four Thirds, MFT, sensor with 2x crop factor)
Now, focal length and aperture are the other two factors (besides the sensor size) that determine how your bokeh is going to look. In our case, we wanted to have a consistent equivalent focal length through all three cameras. A 90mm lens on a 5D series camera isn’t a 90mm lens on an Olympus, so we basically need to make sure that our angle of view is more or less the same. With that in mind, here’s our lens selection.
All three lenses now have more or less an equivalent focal length of 90mm. There are some small differences in actual execution – the aspect ration of the MFT sensor, for example, is 4 x 3 instead of the 3 x 2 of the Canon and Fuji, but they were all shot from the same spot, fixed onto a tripod so that distance between camera, subject, and background is all the same.
The Test Using 3 Sensor Sizes
With that set, we proceeded to take pictures in Aperture Priority mode at every 1/3 f-stop. f/2.8 was the widest that all three cameras/lenses could go, so let’s take a look at that.
Right off the bat, we can see that f/2.8 apertures aren’t the same on every sensor. Shot in Aperture Priority, all three cameras exposed within 1 stop of each other as far as shutter speed goes – though the Fuji likes to overexpose by a stop, which it did here. But the resolution chart in the background rendered completely differently in all three shots.
Now let’s look at them at f/8. Once again, clockwise from top-left: Canon 5D Mark III, Fuji X-E2, Olympus OM-D E-M1.
Again, the three cameras render the res chart in the background very differently. Finally, let’s go all the way to the narrowest aperture all three lens/camera combinations can support: f/22.
The results are exactly as expected, though at f/22 they are more similar than at most other apertures.
So – what did we learn here? Well, we knew that the full frame camera would give us the most blurred and smooth-looking bokeh, and that’s pretty much what we see at all f-stops, from f/2.8 to f/11, to f/22 and everything in-between. We know that the MFT chip, at half the size, doesn’t do nearly as good a job of blurring the background as the full framer or the APS-C sensor.
That much we already knew, but it’s handy to see just what f/2.8 on a full frame camera compares to f/2.8 on an MFT camera in visual terms.
What’s f/5.6 on a Full Frame Equivalent to on an Olympus/Fuji/etc.?
Let’s take this one step further though. We know that f/2.8 isn’t the same in so far as depth of field goes. But what does f/2.8 on an MFT camera equate to on a full framer? How about f/4 on an APS-C?
Well, the first thing to note is that while you can calculate the conversion mathematically, the results aren’t going to be exact. That’s because there are too many variables here – sensor design, resolution, optics, in-body processing pipeline, and more. What we’re trying to do here is to get you as close as possible to some answers about the effect of sensor size on DoF.
The best resource we found for this comes to us via a tutorial on Cambridge in Color. There’s a nifty DoF calculator called “Depth of Field Equivalents” on that page, and we can see that f/2.8 on a full frame sensor is f/1.4 on a Micro Four Thirds sensor and f/1.8 on an APS-C.
Our lenses on the smaller cameras in this experiment don’t go that wide, so let’s start with the widest they can go – f/1.8 on the Olympus 45mm f/1.8 lens.
As we can see, that equates to f/3.6 on the full frame sensor camera, and according to the same tool, that equates to f/2.4 on the APS-C sensor camera.
With that in mind, let’s look at three images with those aperture settings.
Clockwise from top-left: Canon 5D Mark III at f/3.5 (the closes we could come to f/3.6), Fuji X-E2 at f/2.4, and Olympus OM-D E-M1 at f/1.8.
Not bad! The blurred resolution chart looks pretty close in all three image. That Olympus image, though, has a resolution chart that looks a bit… sharper than the rest.
After much experimentation, I happen to know that the Olympus has a tendency to underexpose by as much as two-thirds of a stop. The slightly underexposed image has a tendency to look a bit sharper than it actually is.
Let’s look at one more set. Here we have, clockwise from top-left: Canon 5D Mark III at f/5.6, Fuji X-E2 at f/3.6, and Olympus OM-D E-M1 at f/2.8.
I adjusted exposure on the Olympus file to make it look more like the rest in terms of brightness.
The Fuji file looks just a bit blurrier (again, I know that’s not a word) than the rest, which reinforces my point that the equivalents aren’t exact. But it’s pretty close, and you now have a good foundation to judge the impact of your camera’s sensor size on background blur.
Here’s how to do the math in your head – or in a calculator.
Going from MFT to full frame is easy. Divide the f-stop on the full frame by 2 to get the equivalent depth-of-field (DoF). So, f/8 on a full framer would be give you the approximate DoF of f/4 on an MFT camera.
For APS-C, you divide the full frame f-stop by 1.5 for Nikon and Fuji, and 1.6 for Canon. So, f/5.6 on a Canon 5D Mark III would be f/3.7 on a Fuji, or f/3.5 on a Canon 7D.
The key thing to take away from this experiment is this: If a shallow depth of field is your end-all, be-all, go for a full framer. That said, it’s not impossible to get a decent equivalent DoF on smaller sensor cameras. For example, f/1.8 on a Canon 5D Mark III would be f/0.95 on an MFT camera would be f/1.2 on a Fuji, and f/0.95 on an Olympus.
Pretty crazy numbers – until you realize that both Fuji and Olympus have glass that opens that wide. MFT users have f/0.95 lenses at 17.5mm, 25mm and 42.5mm from Voigtlander, while Fuji has the 56mm f/1.2 R lens for their cameras.
Yes, getting the equivalent full frame DoF of f/1.2 isn’t quite possible on the smaller cameras, which is why we here at BorrowLenses are huge fans of the “right tool for the right job” approach — and why we rent so many different camera and lens models.
I’ll leave you with this impromptu portrait I made of Michio with the Voigtlander 42.5mm f/0.95 at f/0.95. Notice how his ear is out of focus while his eyes are tack-sharp.
The biggest error that someone can make on this discussion is confuse sensor size with photosite size.
Bigger sensor doesn’t makes dof different. End.
Bigger photosites has influence on circle of confusion which has influence on dof perception.
But not necessarily a bigger sensor has a bigger photosite, so smaller sensors can have bigger photosites and has the same perception of dof and bokeh.
I say perception because mathematically dof is another totally different story…
I sugest learn about circle of confusion.
And about all this subject, maybe u can try to translate one of my articles:
From what I understand, your test is almost valid.
Almost because you decided to compare 3×2 with 4×3 sensors by choosing the 3×2 horizontal field of view.
In your images, we can clearly see a lot more information up and down from the 4×3 camera.
And obviously with direct consequences in terms of DOF.
As I much prefer 4×3 images, when I use a 3×2 camera without direct aspect ratio rendering (case of many dslr and, unfortunately, all Alpha 7/9 Sony cameras), I always think in 4×3 and therefore, oblige myself to let more distance between the subject and me to reframe later in postproduction (which is a real pain because my concentration is disturbed).
So, from my point of view, we should always compare FF / APS-C with MFT /digital MF with 2 series of tests.
What do you think?
PD. I had the opportunity to test the GFX 50S. Here we have a real photographic tool with many aspects ratio options like the famous 7×6. This is called liberty 🙂
Excellent comparison – explains it all. One thing to add though: if you can live with manual focus then you don’t need fullframe to get fullframe bokeh: the speedbosters/lens turbo (Metabones, Zhongyi and nonames) will do the same job for you: “compressing” the fullframe lens picture onto the APS-C format – maintaining all the original bokeh style plus collecting the fullframe light onto the smaller sensor (85/1.4 bokeh on Fullframe looks the same on 85/1.4 + Speedbooster on APS-C but the “speed” will increase by roughly 1 stop and 85mm lens remains 85mm lens with APS-C).
I used the Zhongyi Lens Turbo II with Sony NEX and now Fujifilm and picture quality is definitely highly impressive for the Zhongyi at $150 ! Of course native fullframe sensor would capture more details as no extra glass is introduced with the same “uncompressed” lens. But on APS-C the same subject size is even sharper using the Zhongyi than with a glassless adapter as the full glass area is now used but compressed to the smaller sensor – this very clearly outruns the loss caused by the extra glass in the Zhongyi (easy to compare by using a glassles adapter and going further away from the subject). I even had the impression the out-of-focus areas show less lo-CA with the Turbo or the loca are rendered softer, contributing to this improvement.
I can only recommend this solution so much if and only if you are happy with manual focus. It’s a relatively cheap way to avoid the cost of fullframe cameras. I used Minolta MD 50 1.4, Samyang 85 1.4, Nikon AI 50 1.4 and others with the according Lens Turbo version. The master of bokeh – MD 58 1.2 does not excel with the Lens Turbo, I do not recommend this combination but the 50 1.4 on the Turbo will give you a similar bokeh confusion as the 58 with a glassless adapter (but still not as creamy as the Minolta 58 1.2 can get).
It actually is a legit comparison in that if you want the same fov by taking the picture from the same location. Of course the results will be different if you compare the same physical fl lenses on the different cameras but then you gotta stand at different lengths away from your subject (obviously). So for comparing pictures/cameras/lenses you gotta keep it as apples-apples as possible and we can best do this by keepibg fov and distance from subject the same. Sure I can get the same bokeh on my mft em-1 as my canon 6d BUT I gotta stand twice as far away. This test simply shows that if you want a particular fov while standing at the same distance from your subject you are going to get the same “frame” but the background blur and DOF will be very different using the same f-stop.
Dca you are right and also you aren’t . You are right that lens behave same way on all sensors. But if you make a crop from ff picture you will get less points on picture out of focus, so in the end you will get less blurry photo even if physically the object on photo will have same dof. If you want to frame it same you must go away but you retain same count of out of focus points so in the end with same lens, aperture and framing you will get different dof. And that’s all about this myth.
On terms of dof and framing you will get same image on apsc like if you mount 75/2,7 lens on full frame. That’s for 1,5 apsc coefficient.
Thanks Michael, for your information. Accordingly, to use the final test above with the 90mm, f5.6 Canon lens, the 60mm Fuji lens would need f2.5 ( = (60/90)**2 x 5.6 ) to produce similar DOF.
But f2.5 is quite ways from the f3.6 that Sohail used in the actual test.
On the other hand, the photos from the test seem quite convincing that equivalency was achieved with f3.6. And the same is true with the Olympus with its f/n based on linear relation with focal length.
I am puzzled.
Michael A. Knox
Robert is correct. DoF is proportional to f/number, inversely proportional to the square of the focal length and proportional to the square of the focus distance. Hence, shorter focus distance equates to shallower depth of field.
I don’t agree with Dca on point 2. At greater distance, crop sensor should show less bokeh than FF. Here’s the reason: As the distance increases, all things get focused closer and closer to the focal plane. When you focus on far objects like the moon, there is no bokeh to speak of even. So greater distance means less bokeh.
It seems you are alluding to the perspective effect by which the sizes of background objects change relative to foreground objects. But given that the lenses are all set at the same location, what they ‘see’ should all be the same.
The 2 factors that control DOF are the focal length and the f-stop. First suppose no cropping is considered, your bokeh effects would be the same on 2 cameras. Now to crop the FF picture to look like the APS-C, you set the zoom to 75mm (=50mm x 1.5). At that greater focal length, your bokeh would be smoother ( thinner DOF). Conversely, to zoom out on the APS-C to look like the FF, set your lens to 33mm. The result is your picture includes a greater angle of view, but with less pronounced bokeh effect.
The key term: In practice.
The reality is, most crop sensor camera users will never get the dreamy bokeh they see in full frame cameras and full frame camera users will always need to jack up ISO to take pictures with similar DOF as crop sensor users in low light / indoor conditions. Unless of course they use flash.
I use m43.
Dca, this is a very good analysis! I’m curious though about the field of view.
In situation 3, although the framing is same, but since the crop sensor uses a wider lens, will the field of view be bigger? Will I see a bigger and wider background? I’m looking at the pictures in the last comparison above and it seems like I’m seeing a bigger background.
I also agree with Brad and Dca. The sensor size does not affect the amount of blur, as far as you use lens of the same focal lenght. However, since FF cameras demand longer focal lenght lens than cropped sensor cameras to achieve the same image framing, they will, in practice, produce more blured background.
1. crop sensor versus FF with same lens from same distance = exactly same DOF and same background blur. (It’s a tie)
2. crop sensor versus FF with same lens with FF shot at shorter distance from subject so it can achieve same framing as crop sensor = crop sensor has less DOF than FF and crop sensor also has more background blur than FF. (Crop sensor wins)
3. crop sensor versus FF with different lenses from same distance from subject to achieve same framing, FF sensor will have less dof and more background blur (FF wins)
3. Conclusion. The myth that FF offers advantage over crop sensor for creating more background blur is false. Fact: A 50mm f1.7 lens is always a 50mm lens with f1,7. Changing the size of the sensor behind it does not change the DOF!. The only reason crop sensor will have less background blur, and more dof is you use a smaller focal length lens on the crop sensor. Considering a nearly all photographers will frame their subject of interest in the real world, the result is FF, APS-C, and 4/3 can all easily achieve exactly the same amount of background blur.
No the conclusions made from this test are completely wrong.. the only reason the background blur is different is because he forced the smaller sensors to use smaller focal length lenses. 60 and 45 versus a 90mm. So of course they have less background blur from the same shooting distance. Repeat the test fairly with each sensor using the same focal length to frame the subject identically. This means the distance from subject will increase with crop sensors (FF sensor will have to shoot from closer distance to achieve same image) All 3 sensors will have exactly the same background blur as the FF.
bokeh and DOF is a 2 different concept. DOF provided from a lens never changed. however, DOF change on the print size or sensor size. Circle of confusion. CoC…. Also DOF dominantly changed by the objust distance which now being a big matter for the FF, Aps and M43 for the bokeh
What is also driving this, is that in order to get the same field of view with the cropped sensor cameras, you have to use increasingly wider-angle lenses which have more depth of field at the same aperture, as part of the inherent optics of the lens. The opposite is true when using wider lenses – for instance, I use a 21mm/2.8 asph on my Leica M8 (APS-H/27x18mm) which has the angle of view equivalent to a 28mm lens on a full frame camera, but still has the inherent very deep depth of field of a 21mm wide angle.
What happens with the DOF when you mount a Full frame lens on a APS-C camera, lets say: 50mm at f1.8 on a Ful frame, and the same 50mm at f1.8 on a APS-C, crop the FX image to get the “angle of view” of the APS-C and what would be the result? Not being rethorical tho, I actualy dont know.
Vangelis Matos Medina
Why the frame is so different on the 2 cameras 3×2? height of the sensor relate to the tripe head?
Blurier isn’t a word, but blurrier is. You misspelled it on porpoise?