Can You Tell the Difference? 4K Galaxy Note 3 vs. Canon 5D Mark III Video

Iddo Genuth (903542) writes "Photographer and videographer Alec Weinstein was in the market for a new smartphone. He realized that the new Samsung Galaxy S5 and the Note 3 both have 4K video recording capabilities and decided to compare those to his 1080p 5D MKIII pro DSLR camera – the results are extremely interesting — Can you tell the difference between a Canon 5D MKIII shooting 1080p video and a Samsung Galaxy Note III smartphone shooting 4K video?"

Comments

EXIF

[alex4u2nv]

Yes, their exif meta tags are different ;)

DOF

[Teun]

The site is slow to load (surprise?)

I assume the obvious difference is going to be the depth of field or DOF.
The Galaxy will have oodles of it but lacks the ability to isolate the subject, the Canon will make a nice sharp shot on the subject leaving the surroundings vague.

And then there's this thing with zoom/ interchangeable lenses...

Re:DOF

[Teun]

Just saw the video via the Youtube link below.

That phone was awesome!
But I would also like to see some footage out of the sun or on a gray day.

Re:DOF

[Anubis IV]

Indeed. And besides depth of field, putting a 16MP camera in the phone means that the amount of light hitting any particular pixel of the sensor will be ridiculously small, resulting in a reduced dynamic range. That design decision leads to pictures that end up looking worse, though your Average Joe won't be able to tell the difference anyway. Even so, the megapixel game is virtually meaningless for daily use once you get past a certain threshold, and we passed that point years ago, which is why other manufacturers are increasing the size of their pixels, rather than trying to pack more pixels in (Nokia being an exception).

Re:DOF

[viking80]

I respectfully disagree on all your points
- Small pixels reduce sensitivity, not dynamic rage, but the whole point with the isocell sensor is to increase sensitivity in a small pixel. Because photons are discrete, your dynamic range can be no better than 10*log(photon count/pixel). To get 10 bit dynamic range you need 10e3 photons/pixel.
- The megapixel game is not meaningless. I use a large printer, and with a 25Mpix sensor, the result is a lot better than with a 10Mpix sensor. The print actually has a resolution of 12 000 Mpix!

The quantum efficiency, QE, of most backlit sensors ranging from the best DSLR to the Samsung is all around 10%. (Human eye and astronomical cameras can be up to 100% i.e. detect single photon.)

10% QE is about 5 picoLumens per pixel sensitivity, and here is where the sensitivity comes in. 1 lux= 1 EV = 1 lumens/m2 = a bit more than bright moonlight. Assume you have an f1 lens. now you will need 5 nanoLumens/pix for 10bit DR.

A 7mm lens will give you 3.8E15 photons/s, so each of the 16Mpix will get 2.38E8 photons, or 2.38E7 LSB. This should equal 24bit dynamic range. This is with a lot of generous assumptions like an f1 lens, no statistical noise, no thermal noise etc, but still enough photons to give good dynamic range in the darker parts of a photo.

This should give some insight into some of the fundamental limits.

Re:DOF

[asvravi]

A whole lot of hogwash in here - wrong units, dimensionally inconsistent equations, plain ridiculous or missing assumptions but still the post gets modded as insightful just because it *sounds* insightful.

- Larger pixels improve dynamic range. DR is defined as max signal before pixel saturation, divided by noise. Noise is limited by shot noise and electronics so does not scale with pixel size. Larger pixels have more signal range. So DR is higher.
- You calculate DR as if there is only one electron noise. Try several magnitudes higher noise! I am not sure DR is what you think it is.
- QE for most sensors is between 20% and 50%. 10% is nonsense.
- ISOCELL improves color rendition, it has nothing to do with sensitivity.

Following from Samsung should help -

According to Samsung, the ISOCELL sensor design achieves better image quality than is normally possible from the very small CMOS sensors used in smartphones and tablets. ISOCELL uses a backside-illuminated (BSI) photodiode that is unique compared to past designs thanks to its integrated barriers between the individual pixels. Compared to conventional BSI sensors, this reduces electrical crosstalk by about 30 percent. Crosstalk - the bleeding of photons and photoelectrons between neighboring pixels - has been a disadvantage of traditional BSI sensor design, one that can reduce image sharpness and color accuracy because light intended for one particular pixel spreads to its neighbors.

Existing BSI designs, with their photodiodes near the front of the sensor, lack any inherent structures that prevent light bleeding between pixels (a role fortuitously played by the circuitry in front of the photodiodes in older, frontside-illuminated chips). The barriers in the ISOCELL design prevent this bleeding.

How do you equate 10% QE to 5pLumens/pix "sensitivity"? I am not sure Sensitivity is what you think it is. Sensitivity is defined as voltage output from the sensor for a given light input. What is the voltage output assumed here? How does it compare to the camera noise?

Given this, rest of your statements do not make any sense either. When you say "generous" assumptions, it turns out they are actually ridiculous assumptions - you have removed the entire point of analysis and pixel size and even ignored reality, which is what the OP is commenting about. You disagreed with his points that are based on solid reality, but then ended up giving a half-baked proof derived from supposedly "fundamental" limits that are nowhere close to reality.

No different than asking...

[Pollux]

Can Joe Sixpack tell the difference between a $10 glass of house wine vs. a $100 glass of 1982 Chateau Gruaud Larose?

Besides, why would I use a DSLR to shoot video? Wrong tool for the job. That's like using a Ferrari to haul construction equipment or using an F-150 on racing day.

On the other hand, just try to use a smartphone to take pictures of fireworks at night or shoot a picture of your child making a layup at his basketball game in an indoor gym. Then tell me how the two compare.

Re:No different than asking...

[dagamer34]

Are we assuming perfect world or what the average enthusiast might have? Because I'd rather shoot video on a dSLR which has far better optics and real physical zoom than a smartphone that has to cram everything into 7mm or less of space.

Re:No different than asking...

[wasteoid]

Wine snobs are the worst kind of snobs.

Re:No different than asking...

[MichaelJ]

There's no real-time autofocus in nearly all dSLR video recording, including the 5DMKiii used here. In Canon's lineup, only the 70D can do it.

Re:No different than asking...

[Anonymous Coward]

Can Joe Sixpack tell the difference between a $10 glass of house wine vs. a $100 glass of 1982 Chateau Gruaud Larose?

That's a pretty funny example to use because oenophiles can't tell the difference either. There is a HUGE reason that wine tastings are not done blind: it is because the wine experts can't tell the difference. In the 1970's there was an international wine competition done blind, and California did exceeding well. It gave instant credibility to California wines and the French cried foul over the results and the process of the competition (the result was to revert back to knowing the label during the competition). Fast forward about 30 years and another blind competition was done, and "2 buck chuck" did exceedingly well. Of course, the California wineries cried foul over the results and the process of the competition.

Are you a cork guy as well? You do know that screw caps are far superior closures for wine, don't you (as cans are over bottles for beer, and I would LOVE to see wine in cans but can you imagine the ignorant OUTRAGE you'd get from the wine idiots?)?

Re:No different than asking...

[dgatwood]

The capped recording time is actually the fault of the European Union's import duties, which charge a higher tax rate for anything that can record 30 minutes or longer. Blame excessive government bureaucracy for your DSLR being crippled.

Re:No different than asking...

[rogoshen1]

As long as it allows people to take:
obnoxious duck faced selfies
obnoxious pictures of food
cell phone cameras have been 'good enough' for years.

Now when it comes to joe and/or jane sixer taking pictures of 'real important events' (that like 99% of the photos taken, will never be viewed -- ever) does the DSLR's better image quality really matter? Looking back at childhood photos (my parents had a polaroid) I couldn't possibly care less that the photo is a bit grainy, or that the camera didn't take pictures with the sensitivity of a fucking CIA spy satellite.

Re:No different than asking...

[vux984]

Are you a cork guy as well? You do know that screw caps are far superior closures for wine, don't you

I prefer corks. Because I enjoy opening wine bottles with corks. I can't tell the difference in the wine unless its actually spoiled. I know screw caps are better seals but its not as much fun.

(as cans are over bottles for beer,

And I prefer bottles too. Because i like the sensation of a cold bottle on my lips more than a cold can.

Just as I prefer like drinking anything from a glass or mug over drinking it from a plastic or paper or metal cup (whether its water, juice, milk, tea, or coffee...)

I would LOVE to see wine in cans

I'm sure that'd be fine in terms of taste as I'd still drink it out of a glass.

Dining is very much about the taste, but you shouldn't discount the value in the pageantry, theater, and traditions of the experience. They may not affect the taste, but they are still part of the whole experience.

Re:No different than asking...

[petes_PoV]

Actually snob snobs (those who rate snobbishness in different categories) are worse.

Re:No different than asking...

[bugnuts]

why would I use a DSLR to shoot video?

You wouldn't, because by asking this question you betray that you undoubtedly have never shot a video before.

DSLRs have some great features, and potential features if you need them.
0. high quality and cheaper cost than a broadcast quality 2k camera.
1. interchangeable lenses.
2. easy to mount nearly anywhere.
3. large sensor can give a shallow DOF when needed, and decent low-light ability.
4. some can shoot raw footage, when needed.
5. can use comparatively inexpensive vintage lenses.
6. easily maintained and replaced.
7. high enough quality for movies, and getting better.
8. well-supported by 3rd parties.
9. often have very usable ISOs, esp with a little bit of noise reduction (of which there's exactly one good program).
10. have spawned camera offshoots based on DSLR video which is closer to a movie camera/dslr cross.
11. can be operated remotely over usb or wifi. This includes focus pulling.
12. firmware can be hacked on some, unlocking even more features.
13. can be used as a crash camera for larger budgets.
14. can be housed for underwater shooting.

Some of the problems with DSLRs for filming. Not all will apply on any particular shoot.
-1. large sensor can be a big hindrance when you need a large DOF, and requires a lot more light than a small sensor.
-2. most movie modes are afterthoughts. Very few decent still cameras also have decent movie modes.
-3. very few have any sort of usable auto-focus, although some can lock on and track. Autofocus pulling usually sucks.
-4. very few have genlock, SDI, or aux i/o or undecorated uncompressed output
-5. most outputs are in 8 bit 4:0:0 which loses a lot of color information. Some have 10 bit 4:2:2 and this is changing as memory speeds increase.
-6. many don't have a very good codec and bit speed, but this is also changing.
-7. most limit recording to 30 min due to EU taxes. Not usually a problem except for conferences and long interviews.
-8. no global shutter. This is usually a very expensive feature, although at least one offshoot has it for under $10k. Maybe $5k.
-9. limited fps speed adjustment. Some small cameras can shoot up to 1000fps for a short time, but dslrs can't do even a short slo-mo section. Some will do half-speed.
-10. Not as ergonomic as a dedicated movie camera. As a DoP, this can affect things.

All of the above can be found pretty easily if you know what to look for, and that should give you plenty of reasons why it is in many studio's interests to explore what DSLR shooting can bring them. I've shot several shorts, movie videos, and a TV show. Most were with a DSLR.

Re:No different than asking...

[Anonymous Coward]

Whoa, whoa, whoa... what's going on here? A /.'er understanding some of the values of living life? Colour me confused.

Re:No different than asking...

[ignavus]

I would LOVE to see wine in cans

Cans are opaque.

Re:No different than asking...

[Kjella]

I still think the primary reason people use DSLRs to shoot video is even simpler, overlapping jobs and overlapping skill set. I mean if you want someone to make a video of your wedding, you probably want wedding photos as well. Pretty much everything about making a good photo (focus, exposure, composition, lighting etc.) can be applied to making a good video. So when you're thousands of dollars invested in camera, lenses and you know it inside and out, you'll still be a better man on your DSLR than a rental broadcast cam with cine lenses. While still cams can make decent video dedicated video cameras generally can't take stills any professional photographer would want to use, so if you buy one you're deeply committed to being a film maker. Most simply aren't so purebred.

Re:I could presumably count the pixels?

[Jeremiah Cornelius]

ANSWER?

Lenses, lenses, lenses...

OMGPWNIES

[wickerprints]

Let's see if the Galaxy Note 3 can:

1. Record usable, relatively noise-free video at EV -2
2. Use f/1.2 lenses
3. Record at effective focal lengths wider than 24mm or longer than 85mm...how about video at 300/2.8 or 600/4?
4. Use varifocal lenses of any kind, let alone a parfocal lens

I mean, this is silly. Under a very limited subset of possible shooting conditions and configurations, you *might* be able to get comparable output, but this has no bearing on the fact that if you're using a $3000 DSLR to shoot video, you're not merely some Android fanboy taking selfies of yourself beating off in your parents' basement. You're looking at using it with cine lenses or even just EF lenses like the 24/1.4L II, 35/1.4L, 50/1.2L, 85/1.2L II, 135/2L, 200/2L IS, or 300/2.8L IS II (if you're addicted to primes). Or Zeiss if that's your poison. Good luck with mounting a 55/1.4 Otus to that Galaxy Note.

Re:OMGPWNIES

[Rich0]

Let's also see which one fits in your pocket, making it available at all times..

Cell phone cameras have their place. However, if you're actually going someplace to take photos, they're not the tool you should be reaching for.

I have a leatherman for situations where I just need a bunch of tools handy, but when I do maintenance on my car, I grab my toolbox and not my leatherman. Sure, I could probably manage to get the oil plug out using the pliers tool on it, but I own a socket wrench and a set of metric sockets for a reason. When I'm going to change tires, I grab my breaker bar too, and my torque wrench for putting them back on.

That doesn't mean that there aren't situations where the leatherman is handy to have.

Re:OMGPWNIES

[Solandri]

Cell phone cameras have their place. However, if you're actually going someplace to take photos, they're not the tool you should be reaching for.

I have a leatherman for situations where I just need a bunch of tools handy, but when I do maintenance on my car, I grab my toolbox and not my leatherman. Sure, I could probably manage to get the oil plug out using the pliers tool on it, but I own a socket wrench and a set of metric sockets for a reason. When I'm going to change tires, I grab my breaker bar too, and my torque wrench for putting them back on.

It doesn't work like that. Your tires and the lug nuts are always the same size.

With a camera, if you pair a small lens with a small sensor, you can produce the exact same image size upon viewing. So long as the minimum conditions of diffraction/resolution, optical quality, and sufficient photons per pixel to keep noise below a threshold level are met, the images from a small camera (your leatherman) and a large camera (the socket wrench set) are indistinguishable. It's only in the more extreme cases (low light, telephoto) where the larger camera starts to pull ahead.

The images from a modern cell phone camera are competitive with the images from a DSLR from about 15 years ago. Sensor noise has been reduced and sensitivity increased. You can apply these improvements by either holding the sensor size (and pixel pitch) constant and getting much higher ISOs (we've gone from 1600 ISO as a max to 256,000 ISO as a max on DSLRs). Or you can hold ISO constant and reduce sensor size. Cell phone cameras just do the latter. As long as you aren't trying to make a telephoto or shoot in low light, the tiny lens size doesn't hurt you because we aren't yet close to the diffraction and resolution limits for the 20-40mm equivalent that's typically found on cell phone cameras. We still have some size reduction that's possible before we'll hit diffraction limits.

Re:best camera is one on hand when need it

[zr]

direct youtube link to the comparison: https://www.youtube.com/watch?...

Re:best camera is one on hand when need it

[Rich0]

seriously speaking, under good lighting conditions phone cams have been on par with SLRs for all practical purposes for quite a while now.

Under good lighting conditions $20 cameras have been on-par with $800 cameras for decades.

The problem is that good lighting conditions are fairly rare.

I'd love to have a decent phone camera, but it is really hard to accomplish in optics that are a few mm across what you can do with optics that are several cm across. My current phone camera is very prone to lens flares, has fairly poor dynamic range, and isn't terribly light-sensitive. It captures plenty of blurry pixels though.

Someone should kick the guy in the balls

[toxygen01]

This is most likely a promo for galaxy. Aperture and focus were intentionally set wrong so that 5D mkIII looks just a bit worse. marketing at its best.

Re:best camera is one on hand when need it

[Rich0]

Interesting. It looks like the 5D pictures weren't processed at all - they could stand some sharpening.

That is important, because most consumer-oriented cameras (such as in phones) apply quite a bit of sharpening automatically. Professional cameras almost always avoid any processing, so that you don't get further losses when you post-process them (professional photographers almost always post-process).

Also, half of the comparisons used JPEGs taken by the 5D. Again, serious photographers rarely capture JPEGs, because they lose dynamic range and end up being double-processed (and they're lossy besides).

I did note that the RAW photos taken by the 5D handled one of the high-contrast shots much better (lots of shade in the foreground with the sky and sunlit buildings in the background).

Focus speed is also an issue with cell phone cameras - a DSLR will focus MUCH faster and more accurately, especially in low light. They can also capture pictures in rapid succession. Most DSLRs are designed so that if you push the shutter release, they take a picture, no matter WHAT else is going on in terms of modes/etc (well, unless you put it on delay timer or something). A DSLR is always ready to take a picture, and will do so very quickly.

The main advantage of the cell phone is that you always have it on you. However, if you're actually planning on taking photos, I'd pick the DSLR any day. There is just no comparison in the photos they take except under the most ideal conditions.

In a school zone all cars are the same.

[sandbagger]

However, the moment you're doing anything else, the differences show. So, yes, at two paces away in perfect daylight, with no need for special considerations, yes, a smart phone will take decent photos. Given that even at press conferences telephotos and zooms are needed to see the podium, or you're shooting in imperfect light, or you need a polarizer, or you need to add off-camera flash, you'll need a decent camera.

A few years ago, people were saying that new manufacturers would emerge because Nikon and Canon were wedded to an old-fashioned camera format and the multimedia still/video camera would emerge as a new UI. Well, PJs are still shooting with a design perfected over generations and those needing to shoot video bolt the cameras onto harnesses that make the rigs no smaller than Betacams.

Re:I could presumably count the pixels?

[geekmux]

ANSWER?

Lenses, lenses, lenses...

Ah, true, but only half-right.

In today's world, the much more relevant factor is playback, playback, playback...

(In other words, how many of your 100 friends have the hardware even remotely capable of true 4K playback...needless to say, this is a solution without a problem...)

5D III video is inherently soft

[Diamonddavej]

The 5D Mk. III applies a strong low-pass filter after a rough line-skipping down sampling step when transforming an original 21 megapixel image into 1080p video (the Mk. II is worse). This results in soft looking video with a subjective resolution more like 720p than 1080p. It's an unfair comparison.

However, professional film makers that use the 5D Mk. II and 5D Mk. III cameras shoot in 2K and 4K Raw by using Magic Lantern (no in camera re-sampling or low pass filters, just pure sensor data). Magic Lantern is a end user project that has produced an alternative firmware for Canon DSLRs which has greatly extended camera capabilities and video quality.

The results are spectacular:

https://www.youtube.com/watch?...

Re:Uh, we need a new monitor for that!

[petes_PoV]

4K just isn't here yet in monitors.
Then please explain this:

Pretty easy to explain. That monitor is there (on Amazon). I'm here. Ergo, 4k monitors aren't here yet.

When I have a 4k monitor here,in front of me, then looking at 4k video will be a sensible thing to do. But until there's a 4k monitor here it would be pretty pointless.

Re:I could presumably count the pixels?

[sjwt]

Set1) The horrible frame rate and jumpness gave it the first set away, the cheap 4K was jumpy in the redrawing, looked like a series of flashcards, 1080
Set2) I chose the colours and natural look of the 4K
Set3) Zoomed in, I liked the shadows and total rang on the 1080
Set4) Beach - This gave away the cheap 4K easily. Horrible processing on the steps in the background was so distracting I couldn't take my eyes off it. 1080
Set5) Beach2 - Jumpy image was horrible on the 4K, you couldn't use that anywhere you were paid. 1080
Set6) Shado2Sun - The 4K was blown out in the highlights and anywhere close to highlights, 1080

The 4K is a lot better than I was expecting, amazing for the price you buy, but I wouldn't want to use it professionally, you would be called out by someone with 1/2 a clue, and I doubt its going to handle low light levels at all with the amount of blowouts in the highlights, its clearly compensating there.
Try taking the samsung out of the light, and seeing how it goes in say areas lit by 1-3 candles at night, that would be fun!
Canon 5D Mark III with 35mm 1.4L lens low-light noise test

For F16, that seemed to have a horrible Depth of feild on the Cannon, i'm left wondering about that..
This guy seems to have better DOF at F1.8, showing how badly this guy botched up, at F16 everythign should be sharp as a tack unless hes picking an insanely stupid focus point. 5D Mark III Low Light Playground

Let us not forget that what you shoot with a professional level digital camera is supposed to be edited and altered, its captured to preserve as much data as accurately as possible, you then process it.. Imagine running this kind of editing on his video, and this is a MK2 not a mk3
5D Mark 2 RAW Grading and Dynamic Range Test and Graded Da

And last but not lest a Mk3 with Magic lantern firmware and post processing.
5D Mark III 14 bit RAW Video with Magic Lantern

what is this, a samsung advert?

[unami]

obviously, under ideal conditions for the samsung (brightly lit, mounted to a tripod, canon lens stopped down, canon footage graded in post to match the overly sharp look & oversaturated colors of the galaxy), they will produce similar videos. but that's the one exception, not the rule.