Optical performance of the iPhone 8

Courtesy of Apple, I think.

Preamble

I have been silent for months now, my job got in the way of having extra time for writing about strange things, that no one cares about, but me. For example, how good is the optics in Apple’s 2017 flagship model? All these measurements would have been relevant (?) 3 years ago, but 1. I wasn’t around 2. I had no iPhone 8, but an iPhone 6s and a Samsung Galaxy S8.

As mentioned in my previous articles, before you head on to reading, it is recommended to check out the links below. Makes understanding all this stuff easier.


For your convenience

SMC Takumar 1.4/50 on Fujifilm X-E2: An almost scientific photography journey – Introduction a. k. a. Part 1: Introduction
SMC Takumar 1.4/50 on Fujifilm X-E2: An almost scientific photography journey – The setup a. k. a. Part 2: The setup


The setup

The basics are the same as in Part 2, but
1. I have an artificial ~5800 K (measured with X-Rite and DisplayCAL) lighting in a softbox, which gives me the ability to measure even at night and illuminate the target more evenly.
2. I bought a Manfrotto Junior 410 geared tripod head, getting under 1° in any direction just got easier.
3. The Fujifilm X-E2 changed to an Apple iPhone 8, obviously.
4. I used a shitty old chinese phone holder (from a GorillaPod knockoff) on top of the tripod head. The Manfrotto’s bolt was short, it didn’t reach the metal part of the thread, only the plastic one, so it didn’t hold it tight. I had to be insanely delicate while operating the phone. As I said, if it looks silly, but works, it is not silly.
5. The camera software was Adobe Photoshop Lightroom v. 6.1.0 and I used the “Professional” mode setting the color temp, exposure (ETTR with zebras) and ISO. The DNGs were synced to the cloud, then exported as original.


The sensor and lens

As TechInsights published in October 11, 2017,

The wide-angle Sony CIS has a die size of 6.29 mm x 5.21 mm (32.8 mm2). This compares to a 32.3 mm2 die size for iPhone 7’s wide-angle CIS. We can confirm a 1.22 µm pixel pitch.

So the sensor has ~12.2 MPs (12,192,768 pixels), the lens is 3.99 mm (~21.14 mm equiv.) f/1.8 (~f/9.54 equiv.).


Original raw (DNG) and PNG files for download

https://drive.google.com/drive/folders/1G9DsKf2rICnCMTxcNXwd7KnL0hJqtg9Q?usp=sharing


Reading the MTF chart

As Nasim Mansurov states in a PhotographyLife article:

You might be wondering about what numbers in the vertical Y axis can be considered “good” or “bad” for both contrast and resolution. Generally, contrast will typically be higher than resolution in MTF charts, so anything higher than 0.9 indicates excellent contrast, between 0.7 and 0.9 is generally very good, between 0.5 and 0.7 is average and anything below 0.5 is soft / bad. For resolution, these figures are obviously going to be a little lower, especially for wide-open performance. But this is my subjective opinion – the ranges for what is considered to be excellent or average will probably vary from person to person.

As Roger Cicala states in his phenomenal article on the LensRentals blog:

Why do we lens testers give you the MTF50 numbers? Well, the first reason is it’s the default reading in Imatest software, so it’s simple and easy. There’s also the fact that most of the lens testers use it and people like to be able to compare results from different testers, so there’s kind of an MTF50 gentleman’s agreement going on. Plus, 617 different graphs showing MTF Every Frequency at lens locations Everywhere just cause an article to be confusing and chaotic.
But the truth is that MTF50 is a probably the most important overall MTF number when evaluating a lens. MTF50 is has been shown in numerous studies to be the point where humans perceive an image to be “sharp” rather than blurry. That makes sense – it’s basically where the contrast is greater than 50%.
But MTF50 is not the only important number. Those other frequencies give us different, but important, information. Lower frequencies, like MTF 80 to 90, show how “contrasty” an image is. If you photograph large, bold structures, this area of the frequency curve may be more important to you than the MTF50. Numbers like the MTF10 or MTF5 are the absolute resolution limit of the lens. They show what the smallest detail that the lens can possibly resolve is. Anything smaller is just smooth gray blur. Trained human observers and image enhancement programs can actually make out some details at MTF5 in a photograph, but most of us need MTF10. Landscape and macro photographers trying to get the most detail in their large prints might consider MTF10 to be more important, or at least nearly as important, as the MTF50.

I will definitely stick with these definitions.


Alignment

From the MTF Mapper user guide:

Most of the MTF Mapper outputs rely on setting up the chart to be perpendicular to the optical axis of the lens, or put dierently, you want your camera’s sensor to be parallel to the test chart. If your test chart is not perfectly parallel to the sensor, it might appear that, say, the right side of your lens appears soft. You can only really say that the right side of the lens is in fact soft if you have conrmed that your set-up is good.

Absolute success, every angle is below 1°. Well, I could have centered the frame better and backed up a bit to have more rectangles towards the edges. Click on image for full size.

MTF50 datasets

MTF Mapper settings

Settings for MTF50.

Annotated image

From the MTF Mapper user guide:

In an annotated output image, the MTF50 value of an edge is printed on top of the edge itself. These annotated images are useful to evaluate the sharpness of your camera across the entire image, but you want to see the numbers, rather than a plot. The colour of the annotation text is Cyan if MTF Mapper thinks everything is ne, but will change to Yellow (edge orientation is close to 45°, or the edge is shorter than about 25 pixels, which may result in inaccurate results).

Click on image for full size.

Surface images

From the MTF Mapper user guide:

Derived from the same input images as used to produce Annotated images, the MTF50 surface images are a colour representation of your MTF50 values measured across your camera’s sensor. Note that MTF Mapper’s output is split into meridional and sagittal plots; the main reason for this is that most cameras have distinctly different performance along the meridional and sagittal directions. The sagittal MTF50 values are derived from edges that are oriented along radial lines (with respect to the centre of the image), and the meridional MTF50 values come from the other edges, that are oriented to be tangential to a circle centred at on the image;

2D MTF surface images

Click on image for full size.

3D MTF surface images

Click on image for full size. Looking at the meridional lines, I may have a wavy field curvature.

MTF10 datasets

MTF Mapper settings

Settings for MTF10.

Annotated image

Click on image for full size.

2D MTF surface images

Click on image for full size.

3D MTF surface images

Click on image for full size.

MTF90 datasets

MTF Mapper settings

Settings for MTF10.

Annotated image

Click on image for full size.

2D MTF surface images

Click on image for full size.

3D MTF surface images

Click on image for full size.

Lens profile plots

From the MTF Mapper user guide:

MTF Mapper can produce plots that are similar to those produced by lens manufacturers, which show contrast plotted as a function of distance from the lens centre, at a few selected spatial resolution choices.

Contrast of 0.9/0.7/0.5 at the center. Click on image for full size.
Contrast of 0.9/0.7/0.5 at the sagittal border. Click on image for full size.
Contrast of 0.9/0.7/0.5 at the meridional border. Click on image for full size.

Discussion and conclusion

Scope of analysis

  1. When do sagittal and meridional lines reach contrast of 0.9 and how do they translate to cy/PH (cycles (line pairs)/picture height).
    • Center (MTF Mapper’s first data point on the lens profile graph)
    • Sagittal border (MTF Mapper’s last data point on the lens profile graph, ~2/3 of the image circle)
    • Meridional border (MTF Mapper’s last data point on the lens profile graph, ~2/3 of the image circle)
  2. When do sagittal and meridional lines reach contrast of 0.7 and how do they translate to cy/PH.
    • Center
    • Sagittal border
    • Meridional border
  3. When do sagittal and meridional lines reach contrast of 0.5 and how do they translate to cy/PH.
    • Center
    • Sagittal border
    • Meridional border
Image circles. Click on image for full size.

Analysis with lens profile plots

  1. Contrast of 0.9
    • Center: 45 lp/mm, 260.50 cy/PH
    • Sagittal border: 32 lp/mm, 166.72 cy/PH
    • Meridional border: 24 lp/mm, 125.04 cy/PH
  2. Contrast of 0.7
    • Center: 112 lp/mm, 583.52 cy/PH
    • Sagittal border: 78 lp/mm, 406.38 cy/PH
    • Meridional border: 56 lp/mm, 291.76 cy/PH
  3. Contrast of 0.5
    • Center: 173 lp/mm, 901.33 cy/PH
    • Sagittal border: 136 lp/mm, 531.42 cy/PH
    • Meridional border: 102 lp/mm, 291.76 cy/PH

Statistics

MTF50MTF10MTF90
detected rectangles131131131
sum of all lp/mm values75,732.59182,208.8919,543.99
lp/mm/rectangle578.111390.91149.19
lp/mm/edge144.53347.7337.30
maximum (lp/mm)197.98427.7857.64
minimum (lp/mm)95.81252.4820.52
arithmetic mean (lp/mm)143.43345.0937.02
standard deviation19.4331.916.83
median (lp/mm)132.79347.4936.90

Resolving power of the optical system

According to the 2D/3D MTF surface images and lens profiles, the optical performance is best in the center (and first quarter-third of the frame). The sagittal/meridional borders lag behind by 28.89-50.00%.

Sadly, MTF Mapper doesn’t let you export the sagittal and meridional values separately, but it is clear, that the system resolves 20% more at the higher lp/mm values in the sagittal line, than in the meridional line.

At MTF50 the values vary between 95.81 lp/mm and 197.98 lp/mm, which translates to a minimum of 499.17 cy/PH and a maximum of 1245.29 cy/PH.

At MTF10 the values vary between 252.48 lp/mm and 427.78 lp/mm, which translates to a minimum of 1315.42 cy/PH and a maximum of 2228.73 cy/PH.

At MTF90 the values vary between 20.52 lp/mm and 57.64 lp/mm, which translates to a minimum of 106.91 cy/PH and a maximum of 300.30 cy/PH.

I may have a wavy field curvature. And based on the lens profiles, those may show astigmatism in the optical system. I say may, because as Cambridge in Colour states in an article:

With wide angle lenses, M lines are much more likely to have a lower MTF than S lines, partly because these try to preserve a rectilinear image projection. Therefore, as the angle of view becomes wider, subjects near the periphery become progressively more stretched/distorted in directions leading away from the center of the image. A wide angle lens with significant barrel distortion can therefore achieve a better MTF since objects at the periphery are stretched much less than they would be otherwise. However, this is usually an unacceptable trade-off with architectural photography.

For comparison, the Canon EF 20 mm f/2.8 USM at f/8 on a Canon EOS 5DII gives us a minimum of 1401 cy/PH and a maximum of 1769 cy/PH, the Sony FE 20mm f/1.8 G at f/8 on a Sony Alpha 7R gives us a minimum of 1837 cy/PH and a maximum of 2395 cy/PH respectively. The iPhone 8’s optical system won’t be a match to full frame regular and premium fixed optics due to its sensor size (and sensel number), which is no surprise.

For a more fair comparison, the Olympus M.Zuiko 12mm f/2 ED at f/4 on an ancient Panasonic GF1 can deliver a minimum of 1025 cy/PH and a maximum of 1249,5 cy/PH. A sub 100 USD camera with a sub 400 USD lens is more of a match for the iPhone 8. Well, the former setup still performs better at f/2, which is unreacheable for the iPhone 8 in equivalent terms.

But, hear me out! You can easily have the iPhone 8 for under 200 USD and get a potent camera with small footprint and optical image stabilization. And of course it doubles as a phone and a multimedia device. Profit!


Shortcomings of the test and room for error

Is my test setup precise enough? Did I eff it up? Lets see.

  1. I only tested one particular device specimen, so yours may have totally different optical qualities.
  2. I focused at the center of the target and did not refocus on the edges, so we can only know how this device performs, when focused in the center. It is okay with me.
  3. A3 chart printed on matte paper in unknown resolution. Though it looks adequate for me.
  4. Lighting wasn’t really even, it was stonger on the right side. For now I only have one diffused light source (60×60 cm softbox) on the right and an 80 cm reflector disc (silver) on the left side of the target.
  5. I am no optical engineer, I may misinterpreted my findings. I constantly seek for information and learn as much as possible. Working on getting better.

Final words

This was fun and I still have a few ideas to write about. Definitely want to try if Adobe Photoshop Lightroom CC’s lens profile and Enhance Details option gives any advantage over the raw DNG. I also would like to test if sharpening has a positive effect on the numbers (spoiler alert: it does).


Next up: Optical performance of the iPhone 11 Pro

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s