mrschmo: Can you do depth screening with a Lytro Illium camera for stills?
What good is “the face in focus” if diffraction restricts the total image resolution to 1.5MPix equivalent?
Did you actually USE the system to achieve the effect you describe?
nathanleebush: You guys were much too delicate with this monstrosity. "Not so good for: photographers who require RAW file support" .. What is this, 2003? How did the A6300, which seemed to fall from the heavens, only score 5 percentage points higher? It's half the price with 10x the features, and a nicer image to boot, for both stills and video! This emerged like a hot turd from the depths of hell, and will soon return whence it came.. Canon really needs to get it together, but they obviously are not aware of the internet, or they would have fixed their broken product development culture by now. It's amazing to watch them blow their massive video lead with disappointment after disappointment.
This is supported by his claim that ISO800 is not good, but ISO100 and ISO200 are acceptable — with (practically) ISO-invariant hardware!
The only sensible approach is, IMO, to ignore ISO setting, and shoot at a particular f-number/angle with a particular headroom…
[However, keep in mind that I have practically zero experience with videography, only with photo.]
I do not think the tests of News Shooter make a lot of sense. Take the “scientific approach” to rolling shutter: using a variable-speed fan without mentioning WHICH speed was selected!
Anyway, while THIS test is not “very scientific”, it addresses a REAL concern: rolling shutter is not good. The other two complaints (aliasing, and noise-at-midgray-level) just do not make ANY sense at all.
Aliasing: the main mode of usage “for broadcast” (shoot at 4K, postrocess to HD) was not addressed at all — and judging by the provided samples, it would COMPLETELY eliminate anti-aliasing.
Noise-at-midgray: with Bayer sensors of today THIS should depend on the area of the sensor ONLY. AND: the measurements make NO SENSE at all unless one specifies the headroom available for “brighter-than-white”.
So my first idea is that the guy just does not know that different systems need different level of EV-correction to achieve the same headroom…
yslee1: Sad. It's usefulness in stills might have been questionable, but in video it would've been awesome. Do your focus pulling during post, instead of the hectic nature of the live shoot.
Maybe… But remember that what we discuss is “blurring”, and not “making more sharp”. Plenoptical is always going to be NOT LESS BLURRY than the corresponding “ordinary-focal” shot.
So the possible benefits are not • to shoot something with AF, then “improve sharpness”, but • to shoot something with AF, then “blur selectively”.
1st, I remembered wrong: 1.5MPix limit is for 50° field of view, and (corrected!) 1m–∞ refocusing ability.
2nd: I assume a camera which allows 1Dioptry refocusing ability (e.g., captures image with 1.5MPix resolution at 2m, but allows resolution better than 0.75MPix in the range 1m–∞).
3rd: these are THEORETICAL limits (very similar to Heisenberg principle — and actually proven the same way) for synchronous plenoptical capture. With any particular implementation, one gets even worse results.
Everlast66: "we were competing in an established industry where the product requirements had been firmly cemented in the minds of consumers"
So its customers' fault for not realising the great potential of your product?If I understand correctly, they introduced their innovative product, superior to everything else at same price-point, or maybe solving customers' "big focusing problem", and we customers, with our backward cement-brains simply could not grasp their grand idea?
Sorry, Lytro, but this is totally the wrong attitude!Every decent entrepreneur knows that you can not argue with the market. You have to read it, understand it, and then add something that solves REAL problems better then competitors.
I'd applaud every innovator, even if they fail, but take the lesson out of that failure. But Lytro don't seem to be willing to admit their mistakes and are not showing they learned their lesson, despite its worth of $50 million of investors' money.
> Why hold up today's product as representative> of the potential the technology holds?
This technology holds NO potential. A simple calculation shows that the whole idea is a snake oil. One cannot get more resolution that with a hyper-focal shot by the usual camera. (Read: very blurry, about 1.5MPix max for typical examples of usage.)
YES, there is A benefit comparing to a hyperfocal shot: one can make this blurry image yet-more-blurry away from a particular “focused surface”. But THIS benefit is not what the advertisement were promising.
Depends on the range of the distance you want to be able to focus-pull to. With a reasonable range (IIRC, 2m–∞) the theoretical max of resolution is about 1.5Mpix (with std focal length). Not enough for video applications.
If one restricts this range of distances (say, to 2m–2.5m), then one can reach the resolution needed even for 4K video. I’m not in a position to judge if such tiny freedom in focus post-pulling is worth the troubles.
greenlens: Yes, LoCA is there but no onion rings so far.
I have doubts about LEDs as “bokeh subjects”: too large. For my testing, I reflected LED from a ball bearing. With LED about 10 ball’s diameters away, the reflection is really dot-like.
Could you please fix your reference to read noise of 1e⁻?
You forgot to read http://sensorgen.info/Calculations.html. Summarizing:
The quoted numbers are obtained by fitting a certain formula to the measured data. The formula assumes a simple model of the sensor readout. This simple model works remarkably well for most of the sensors — but it does not work AT ALL for sensors “changing the layout” depending on the ISO setting.
The A7S (ii?) sensor uses a trick (capacity change or somesuch) somewhere about ISO6400 (do not remember the details; see the Jim Kasson’s detailed investigation). As a result, the numbers you refer to are entirely a red herring.
AFAIK, the best sensors have the best read noise close to 2e⁻. (SOME of the numbers on sensorgen which are below 2e⁻ contradict my [much more advanced than curve-fitting] research. So I assume their approach is flawed in this regard too. For example, AFAICS, they do not take into account the quantization of the noise by ADC.)
I’m completely confused by this interview… Correct me if my reading is wrong: He says: • Cinematographers react quickly to changes in tech; • Until the last year, he thought that film has better DR; • AFAIK, in Super 35mm film vs FF digital, w.r.t. DR, digital wins for 8–10 years already (at least for photo cameras).How to combine these? Are cinema camera so far behind?
I see that Alexa is about 6 years old — and it looks it can easily match the performance of photo cameras… How to understand all this?
Onur Otlu: "..and features a new XA (extreme aspherical) element which has been rendered even more aspherical in this lens by being cut in half."
This was my favourite of them all - along with "ugly bokeh".
Thanks for the humor, and Rishi was right - you've *got* to keep this going!
1-element lens is very easy to design — at least if you have a material with VERY high refraction index not depending on wavelength.
The best lens is going to be very thin, going along the sphere. Radius = focal length, center is in the center of image.
This design has f-number up to f/0.5, no spherical aberration, and no coma — but it has very strong astigmatism: the tangential/saggital curvature of field are: flat vs focal length.
From this description, it is very easy to calculate the ellipses of confusion. So this design may be good for supertelephoto only…
Everlast66: I don't know why, but I find this old Nikon F and Canon F-1 much more nicely engineered.Somehow they look much more clean, solid and passionately designed. Of course to some extent this is due to the actual technology being much simpler.
Ha! What do you know about Americans?
Big percentage of them has 6,000 years of history!
YTY: Pretty sure that the Sony 70-200 F2.8 has 64 elements (32 x 2) when cut in half (instead of 46 written in the article)... unless my brain is cut in half and I can't do math.
Looks like it was…
It is 46. (Same misprint repeated again and again…)
ilza: For people who cannot read the original paper:
They have the “lenslet” size Δ=30µ, which, with distance d=500µ to the sensor, gives the blur of 4° (see p.5; I’m using THEIR notations). So, with “normal lens’” 50° field of view, this gives a resolution bounded by about 12×12 pixels.
(They do not state the field of view, but on illustrations, it is close to that.)
Enough said. (I expect the resolution may be improved a bit. I would think that 50×50 pixels may be — theoretically — achievable.)
I’m starting to think that you are missing a lot. First of all, diffraction does not care what is behind the entry pupil: when the incoming rays are scrambled, there is no way to unscramble them. Second, the smaller the pinholes, the WORSE is the resolution (diffraction-wise).
For example, with 1µ pinholes, and 60° field of view, there is non way to get the resolution more than 4×4 pixels (maybe even 2×2 — I’m always forgetting what the coefficient is!).
DP Reviewers, could you please clarify the water-depth rating?!
There are two conflicting standards used in the industries: https://en.wikipedia.org/wiki/IP_Code#Liquid_ingress_protection https://en.wikipedia.org/wiki/Water_Resistant_mark
So, is this 20m rating the IPx8 rating?
Diffraction ALWAYS matters (with small apertures). It does not matter what you put BEHIND the pinholes; it is enough that diffraction scrambles the rays of the incoming lightfield.
With a 30µ pinhole, you CANNOT get resolution above about 1° (with visible light).
About your illustration: can you please mark the page/column/paragraph of the Arxiv paper it would match? I see no relation of your picture to what the paper discusses.
(BTW, I did not look VERY careful into the paper. It looks very similar to the “bright 3D display” idea mentioned here some time ago — and THAT paper I did read through [it was fun!]. If so, your picture has NO RELATION to the idea of the device.)
ms18: Amazing.... And sigma please make some primes also. f/0.95 may be? Even f/1.2 is ok but make sure they sharp wide open as Fuji lenses.
400mm f/5.6 which is far light weight than existing full frame EF 400mm f/5.6. Or increase the aperture to f/4 and keep same weight.
?! The tech was available for several years already.
The only difference now is that it is not built out of two subunits…
Well, I did not try to redesign their system myself — so I’m not 100% sure about the effects of diffraction in their design.
In this sensor, the image at every sensel is formed by a COMBINATION of the individual “squares” of the mask. From this, it is clear that the resolution cannot be better then the blur created by every particular square. However, the contribution of diffraction may be somewhere “in between” the contribution of an individual square, and the contribution of the full mask. (This may also depend on the used algorithm of de-masking.)
So I would not be surprised that the contribution of diffraction may be 2 (or 3? probably not stronger than this) times smaller than what an “individual masklet” would create.
But I already took this into account in my estimate of 50×50 resolution! (And, in their particular design, the diffraction is way below the defocus-blur, so these fine points do not matter.)
No, it is 12×12 image (which is subsequently mutilated by their complicated mask) read by 512×512 sensor — with attempt to demutilate it in post-processing.
The resolution limit is hardwired into the geometry of the mask element. With higher-resolution sensor one might be able to implement a (slightly) better demutilation — but there is no way to reduce the blurring due to the “defocus+diffraction”.
For people who cannot read the original paper: