iPhone 5s Studio Comparison

A few other sites have conducted low-light comparisons of the 5 and 5s and there is a definite improvement. This scene doesn't tell the whole story.

If you don't see any improvements, you haven't been paying attention.

wow, I did not know the GS4 did not have an AA filter? if so then, does it suffer from Moire effects more than say the Z or does the Z also not need an AA filter. I know some Panasonic m43 cameras do not have an AA filter, but in doing so has a software processor to remove moire effects. It seems to work quite well, but still will not remove moire like an AA filter

I don't know whether the AA filter is indeed fully missing or is "only" too weak.

Basically, the GS4 has VERY good effective resolution - and also a lot of (in videos) aliasing / (particularly in stills) color moire. Aliasing is certainly visible for example in the ISO 12233 video framegrabs over at GSMArena. Aliasing isn't really present in double-res ISO 12233 stills (including the vertical double-res resolution checker lines of DPReview's studio scene above) - but color moire is much more present than on the 5s or the 1020.

In the case of the 1020, the video recording and alos the stills, as long as it has pureview 5mp as the output, there should be no moire, as moire is a side effect of bayer style sensors. Because the 1020 can combine 7 pixels into one creating a pure pixel, there should be absolutely no moire, even if the 1020 has no AA filter.

Do you really see the GS4 as having very good effective resolution? I think even though the charts may indicate good resolution, in real life, the s4 images lack textural detail, and also suffers from a massive edge sharpening phenomena.

So is the Samsung S4. They look like coming from the same sensor, watercolor with over-sharpening of edges, just like a normal phone camera.

limlh, the Z has way worse IQ than the GS4. There is just no comparison.

the Z only has worst IQ than the GS4 due to poor image processing algorithms. This is strange since Sony is known to have excellent image processing in their dedicated digicams.

It is said that both the Z and the SG4 have the same 13mp sensor, but the optics may be different and the way the image is processed may be different.

(contd. from above)

Furthermore, if the background were defocused, the 5s couldn't have beaten the 5 in the three-color text test (my dedicated post & crops at http://www.dpreview.com/forums/post/52204330 )

All in all, the 5s is in focus. The corner softness is because of the somewhat wider (30 vs 33mm) & brighter (2.2 vs 2.4) lens and 15% bigger sensor area. It'd be impossible to reuse the same (-size) lens of the 5 with the same corner sharpness. Even Nokia couldn't do it while miniaturizing the 808's lens for the 1020.

"Lumia's superiority is obvious only at high iso, even at 8mp the lens has serious light fall off at corners!."

???? you mean corner softness & CA, not vignetting, do you?

Nevertheless, the 1020 indeed has bad corners / left&right borders. However, in everything else, it wins. Including shooting in bright light like the above test scene (ISO 100 & 1/120s means bright scene): when downsampled to 8 Mpixels, it has considerably less noise than even the (also-downsampled) GS4 or the 5s, let alone the worst of the bunch (the iPhone 5).

Again, feel free to read my dedicated writeup here in the DPR forum, where I've proved this all: http://www.dpreview.com/forums/post/52202175

@KariIceland Yes, sizing down to 8Mp will not only reduce noise, but also reduce the corner softness, but as Menneisyys has stated, that even at the reduce 5Mp pureview, the softness in the corners is still of concern, just less of a concern.

I also find it annoying there is actually no option to do 8Mp oversampling with the 1020, so you either are stuck with the massive full resolution image, or a much lower resolution 5MP image.

I tend to agree that the iPhone and SGS4 do have better corner performance, but one has to remember that the iPhone does have a narrower focal length, which in turn makes it easier to to achieve more consistent sharpness in the corners.

The 1020 has a very wide FoV compared to the iphone. Seeing as size was clearly important to nokia I can see why they went with smaller lenses. Besides, the intention was never to use the full 38MP images, but rather the oversampled 5MP images or even zoom crops from (hopefully) the centre regions of the sensor.

I think the 1020 should include an 8MP option like the 808, and I think this is planned for a future update. In fact I'm sure I read somewhere that this is coming, but I can't remember where. 8MP would be nice because you do get that extra resolution, but you also get the amazingly sharp and noise free image that the oversampling provides. Even the 5MP images from the 1020 have way more detail and less noise than the S4/iphone images, and they often look better than the full 38MP image anyway, so I tend to stick to using those for sharing/publishing.

It would seem that we ought to be able to compare 'it' with some 'real' cameras as well as Vlad has suggested.

Phones are not dedicated cameras yabokkie and others suggest and have little flexibility in some many instances

Have to disagree. The Iphone 5 looks better. Compare the cards and colour circle towards the top right. The card next to the Queen of Spades looks cleaner and better defined on the Iphone 5. Look again at the threads. The Iphone 5s has slightly better thread definition, but the colours are off, with red looking like orange (skin colour on the unshaven caucasian is also orange) and the black-white inscriptions are clearer on the Iphone 5. So, definitely not an improvement - objectively speaking.

I would actually like to see this as well. If we consider phones to be equivalent as point and shoots, I'd like to be able to compare relative performance. I think it makes a lot of sense if I, as a consumer, can decide if the improvements on mobile are good enough relative to a dedicated camera to make either purchase or just stay with one for photographs.

I think you're making a false assumption about the "wide open" lens. Basically, all conventional wisdom is off the table here, since this lens doesn't stop down at all, controlling exposure only through digital ISO and shutter speed. That means it's optimized for the only aperture it has -- there really is no concept of "wide open" because there is nothing else.

That's not to say it isn't impressive, of course!

it's an f/8.5 lens in term of 35mm format, piece of cake.

some complains about peripheral aberration but quality at the center is much more important especially for digital zoom. it's very impressive for it's a lead camera of a new breed of cameras we are going to have, even for interchangeable lens ones because optical extenders will always perform less as well as digital zoom as long as we have enough sensor resolution.

"it's an f/8.5 lens in term of 35mm format, piece of cake".

I think there is a confusion here. In terms of control of aberrations, an F/2.2 lens is an F/2.2 lens, no matter the size of the sensor. Designing an F/2.2 is not a "piece of cake". By the contrary, it is very difficult to control the aberrations, especially in the corners.

well "how difficult" is not something I can calculate with high accuracy. it may be more difficult than f/8.5. still f-number equiv. is a good rule of thumb for cost unless we get really close to f/0.5.

The lumia can resolve a lot of detail but the general performance overall is still heavily in favour of the mid to higher end cameras.

I can appreciate the detail here but it also misses in other areas of scene reproduction.

Eh no, no it does NOT. What you just said is 100% BS.
And I say this as a future owner of that phone (coming in a few days time) But sir you have proven that you have NO knowledge about photography so please just stop claiming things like this, it is utter nonsense.

In the right situation, the Nokia 1020 and it's predecessor, the Nokia 808, would out resolve the 5d3 for detail. I have tested this on the field, and it is the case.

This does not mean the 808 or 1020 is a replacement for your DSLR. There are many other factors that affect capturing images, eg dynamic, range, usability, ergonomic grip, focus speed, high ISO performance, etc. All of these a DSLR will easily outperform the 1020.

If you primarily capture landscapes and use a tripod often, the lumia 1020 or 808 will capture an image with substantial detail but not without some digital noise.

The optics on both the 1020 and the 808 is really quite amazing being able to resolve good detail at full resolutions without suffering from too much diffraction effects. The 808 goes that little bit further by having really impressive detail throughout the frame including the corners, while maintaining an aperture of f2.4.

808 > 1020 in terms of image IQ and overall lens performance..

agree vlad0 I still believe the 808 was designed to compete against large sensor DSLR's whereas the 1020 is designed to compete against entry level point and shoot and other mobile cameras.

there's no evidence that the designer of the 808 targetted DSLRs. the pureview technology in the 808 was designed to be able to perform lossless zoom for pictures and video like an entry-level P&S without making the phone bulky. the exceptional image quality at 1x was just an added benefit. read the white paper if you have any doubt.

It's NOT. Resolution clearly lags behind other 13MP crops of cameras, and has NO manual control whatsoever, lacks OIS.

"I checked them all very carefully and the sponge, brushes and other items look terrible from the Nokia."

If you meant the sponge and the right brush in the bottom right corner, you're right. They're already in the "red zone" of the 1020 - with a native 4:3 shot, the rightmost 6.66% of the frame, that is, 7136 * 0.0666 = the rightmost about 475-500 pixels. Everything in that area (and on the left, of course) is blurred and packed with CA.

However, the center of the frame (about 7136 - 2*500 pixels) has much-much better resolution than that of the 5s. Try, for example, reading the text in the three-color English text, with decreasing character size, in the top center! You'll be able to read almost everything on the 1020, while, on the iPhone 5/5s, only the first few rows.

At least Jonny Ive is a Brit.

empty anti-American comments make me laugh. Americans buy the best product at the best price, regardless of who made it, because that's the American way and that's completely FAIR.

You must be a true Brit and buy all your products from British manufacturers. Which means you have a Triumph motorcycle and not much else :P :P :P

http://en.wikipedia.org/wiki/Manufacturing_in_the_United_Kingdom

yes the corner softness is quite apparent in the 1020. What is interesting though is the SGS4 has the best corner sharpness to resolution, with the older iPhone 5 coming in second. It seems if you want detail across the whole frame, it is easier to achieve using a smaller sensor

"On 1020, 16:9 is already cropped (removes top and bottom of the sensor), the you ask to remove left and right, that means you're using the same size sensor as others"

Wrong - the DPR shots are using the native 4:3 mode of the sensor, which, just like with every multiaspect sensors (also see Pana GH1, GH2, ZS3 and the like) uses far more rows as the native 16:9 mode.

If DPR used the 16:9 mode and, then, cropped to 4:3 (this is what I've recommended and you've also assumed), the outermost 10% would have completely gone. With the native 4:3 mode, 6.666% of the outermost region is still blurred.

Finally, if you shoot 16:9 and crop to 3:2 (it already gets rid of the blurred areas - you don't need to crop to 4:3), you still have much more pixels (and, consequently, much better IQ than even the 5s / GS4) and a comparable FoV (around 33mm equiv - just like with the iPhone 5 and 5s). That is, cropping 16:9 shots to 3:2 is a usable way of getting high-quality images w/o blur / CA.

"DPReview now cut off the left and right of 3:2 sensors (88.9% left) before comparing with 100% of 4:3. "

Nope, the *native* 4:3 mode of the 1020 shoots 7136*5360 images. (The sensor being a truly multi-aspect one, the 16:9 mode is of resolution 7712×4352. This is why there is a considerably wider blurred area when shooting 16:9 - all horizontal photodiodes are used in 16:9, unlike in 4:3.) No cropping was done.

@Menneisyys,

"DPReview now cut off the left and right of 3:2 sensors (88.9% left) before comparing with 100% of 4:3. "

would you kindly read it again please.

we should use each of the target rectangles for each aspect ratio. how the maker makes use of a sensor we have no control but we can test them on level ground and whatever the result should be fair.

we should not cut off a sensor in a test but if that's not possible for sensors of different aspect ratios, we can cut off the same portion of area of each and compare, like if we compare 86.6% of two sensors and decide they are the same, we can say 100% of them are the same, too (not the lens).

DPReview now compares 88.9% of a 3:2 sensor against 100% of a 4:3 one so that if we see two sensors the same in such a setting, we can say 100% of the 3:2 is 0.17 stops better than 4:3.

the same applies to 16:9 only the error introduced by test design may be different.

"DPReview now compares 88.9% of a 3:2 sensor against 100% of a 4:3 one so that if we see two sensors the same in such a setting, we can say 100% of the 3:2 is 0.17 stops better than 4:3."

OK, now I understand what you meant.

In this case, this didn't cause any problems as all smartphones, except for the multiaspect 1020, are natively 4:3; that is, extreme (true) corner softness can be directly evaluated and compared.

(The 1020, while its 4:3 mode uses about 400 pixels less than the 16:9 mode, used its native 4:3 mode as people would do. That is, the comparison to it was also valid.)

Of course, with natively 3:2 sensors (those of DSLR's) corner sharpness can't be evaluated / compared based on 4:3 crops.

I think the same... But pay attention to the ISO. The 5s is shooting at almost twice the sensitivity. I'm not sure why that is so. Did Apple silently raise the base ISO?