ProfHankD

ProfHankD

Lives in United States Lexington, United States
Works as a Professor
Has a website at http://aggregate.org/hankd/
Joined on Mar 27, 2008
About me:

Plan: to change the way people think about and use cameras by taking advantage of cameras as computing systems; engineering camera systems to provide new abilities and improved quality.

Comments

Total: 1187, showing: 1 – 20
« First‹ Previous12345Next ›Last »
On article Sony FE 100mm F2.8 STF bokeh demystified (329 comments in total)
In reply to:

ProfHankD: Just a little note: this lens basically uses the same trick used in the old Minolta/Sony 135mm STF A-mount lens. It's worth noting, however, that the apodization element just applies the gradient to the OOF PSF -- the lack of vignetting (i.e., OOF PSF are circular, not deformed off axis) is because this lens was designed not to vignette, which is a neat trick in itself. Making PDAF work with this is a rather special trick and a big improvement on the A-mount 135mm STF.

mgrum: if it just didn't vignette at f/4.5, you'd see hard edges on the off-axis OOF PSF, which we don't. This has near zero vigetting at f/2.8 (which is T/4.5).

Link | Posted on Feb 20, 2017 at 16:12 UTC
On article Sony FE 100mm F2.8 STF bokeh demystified (329 comments in total)
In reply to:

ProfHankD: Just a little note: this lens basically uses the same trick used in the old Minolta/Sony 135mm STF A-mount lens. It's worth noting, however, that the apodization element just applies the gradient to the OOF PSF -- the lack of vignetting (i.e., OOF PSF are circular, not deformed off axis) is because this lens was designed not to vignette, which is a neat trick in itself. Making PDAF work with this is a rather special trick and a big improvement on the A-mount 135mm STF.

mgrum: in theory, it's not hard to avoid vignetting... but I've got a fleet of 135mm lenses and none of them clip as little as what I see in the sample shots here. Clearly, this was a design priority. It might not just be about having pretty bokeh; it also could be part of why PDAF can be effective despite the apodization.

Link | Posted on Feb 20, 2017 at 13:10 UTC
On article Sony FE 100mm F2.8 STF bokeh demystified (329 comments in total)
In reply to:

Ultan: Why don't they put a transparent grayscale LCD where the iris would otherwise be? Then you could set not only f-stop but also the gradient width between the central clear aperture and the opaque surrounding region. Anybody who wants to patent it, be my guest, but contact me to put my name on the application.

I'm easy with terms, up-front fee $10K, $10 royalty per unit, or let's talk.

Ultan: As I've discussed many times before, the best solution is probably dynamically varying the aperture -- like Minolta itself did in the Maxxum 7 STF simulation mode 25 (I'll never understand why they buried that function so deeply). Anyway, take a look at my post on this: https://www.dpreview.com/forums/post/59134249

Link | Posted on Feb 18, 2017 at 15:27 UTC
On article Sony FE 100mm F2.8 STF bokeh demystified (329 comments in total)
In reply to:

Ultan: Why don't they put a transparent grayscale LCD where the iris would otherwise be? Then you could set not only f-stop but also the gradient width between the central clear aperture and the opaque surrounding region. Anybody who wants to patent it, be my guest, but contact me to put my name on the application.

I'm easy with terms, up-front fee $10K, $10 royalty per unit, or let's talk.

Actually, I've tried this (with a color LC panel). Long story, but unfortunately disappointing results. It turns out that even the slightest texturing shows up very clearly, and LC shuttering has other issues too (remember that LC uses polarization). This is why the Minolta solution (inherited by Sony) of bonding plano-concave smoked glass with same-index plano-convex glass to make an apodizing optical flat is really quite brilliant.

Link | Posted on Feb 18, 2017 at 03:47 UTC
On article Sony FE 100mm F2.8 STF bokeh demystified (329 comments in total)
In reply to:

ProfHankD: Just a little note: this lens basically uses the same trick used in the old Minolta/Sony 135mm STF A-mount lens. It's worth noting, however, that the apodization element just applies the gradient to the OOF PSF -- the lack of vignetting (i.e., OOF PSF are circular, not deformed off axis) is because this lens was designed not to vignette, which is a neat trick in itself. Making PDAF work with this is a rather special trick and a big improvement on the A-mount 135mm STF.

Rishi: The on-sensor PDAF probably does leak like a sieve, so there is more processing going on and the brightness differences induced by the STF are probably made negligible by preprocessing of the signal. I suspect the phase matching is done on the derivative of the signal, which would largely correct for both the leakage and STF shading.

BTW, I'm really amazed that my Canon 5D IV can PDAF with the dual pixels as well as it can because the dual pixel raws confirm that fairly heavy anti-aliasing is corrupting the data pretty significantly. I think it makes a lot more sense to drop the anti-alias filter in main-sensor PDAF cameras....

Link | Posted on Feb 18, 2017 at 00:39 UTC
On article Sony FE 100mm F2.8 STF bokeh demystified (329 comments in total)

Just a little note: this lens basically uses the same trick used in the old Minolta/Sony 135mm STF A-mount lens. It's worth noting, however, that the apodization element just applies the gradient to the OOF PSF -- the lack of vignetting (i.e., OOF PSF are circular, not deformed off axis) is because this lens was designed not to vignette, which is a neat trick in itself. Making PDAF work with this is a rather special trick and a big improvement on the A-mount 135mm STF.

Link | Posted on Feb 17, 2017 at 19:16 UTC as 68th comment | 6 replies
In reply to:

Mssimo: What camera would you buy for $750/$900 w lens?

Well, for $650 you can get a new Sony A6000 + 16-50mm. The two-lens kit including the 55-210mm is $800. I know, that's not a fair comparison... but Canon doesn't make any APS-C sensors as good as the one in the A6000. ;-)

Link | Posted on Feb 15, 2017 at 15:50 UTC
On article Sony SLT a99 II Review (1481 comments in total)

I have a 5D IV. I don't see how this Sony can get lower marks for build quality or video, or a lower score overall. Yeah, all these cameras are quite good, but I don't see how a camera with (by a significant margin) the best FF sensor and a variety of superior specs can earn a longer list of "cons" than "pros" and a relatively low rating... not to mention being reviewed a year after release. It's kind of like reviewing Superman's abilities and saying "yeah, he's super, but he always wears the same awkwardly colored outfit and what's with that cape? No style sense -- 85%."

Link | Posted on Feb 15, 2017 at 13:08 UTC as 81st comment | 2 replies

I really hate that a camera targeted at "entry-level photographers" starts at the body-only price of $750. I know that's partly due to lower expectations for DSLR sales volume, but pricing helps cause the lower sales volume too. By now, I was expecting to see FF bodies in that price range....

Link | Posted on Feb 15, 2017 at 12:36 UTC as 7th comment | 2 replies

It's a strong photo, but of limited "journalistic importance" and I certainly agree that inflating the importance of it sends a bad message. In any case, this was just a contest.

Link | Posted on Feb 15, 2017 at 01:11 UTC as 35th comment

Meh. I think the problem is that my favorite films aren't around now and digital has gotten VERY good. For critical stuff, I used to shoot Panatomic-X @25 souped in 1:100 Agfa Rodinal for B&W -- it told me Tri-X 400. For color, I used to shoot mostly Kodachrome 64, and it suggested two films I never heard of. Difficult to judge anything by little images in a WWW browser....

Link | Posted on Feb 10, 2017 at 02:17 UTC as 18th comment
On article Google AI adds detail to low-resolution images (148 comments in total)
In reply to:

Matthew Miller: > Eventually we might even be able to extract high-resolution images from low-quality security-cam footage a la CSI.

This is an incredibly dangerous line of thinking. These images are not reconstructed — they are _imagined_. It's even clear in these examples: look at the eyelines in the "ground truth" images vs. those in the "upsampled" ones. The constructed images clearly show the subjects looking in a direction they were not. They are, as apparently we say these days, alternative fact images.

The technique might be good for filling in backgrounds or for artistic purposes, but as photographers and technologists we have a responsibility to make it clear that they bear no relation to reality.

It's been easy to create a high-res version of an image for several years, for example, enhancing the image of a car license plate from just a few pixels. The problem is that the text on the license plate would match whatever you used in your training set -- not the one photographed. ;-)

Link | Posted on Feb 9, 2017 at 00:42 UTC
In reply to:

pkcpga: Awesome technology, can't wait to see a larger version. This would definitely help Sony phone sales, only issue is how much memory would your phone need for even just a few short films at 1000 fps and how long can it shoot continuously for. Can't wait to see the final product.

Revenant: I think this is a new fab process, but implementing the same functionality. Still a noteworthy step forward.

Link | Posted on Feb 8, 2017 at 11:21 UTC

#28 looks like what would happen if a clear-bodied Prius drove into a white dropcloth at 70MPH. Anyway, a car seems a bit off topic for DPReview... until one realizes that Toyota already ships an advanced electronics package that uses a bunch of cameras to create a stitched aerial view of the car and surroundings. Come to think of it, why hasn't DPReview done an article on that?

Link | Posted on Feb 8, 2017 at 02:13 UTC as 2nd comment | 1 reply

A new STF lens that can PDAF -- that's a new trick! This is a lens that Canon and Nikon can't touch -- like the Minolta 135mm STF, but AF with OSS and apparently also close focus (1:4). Wow.

Link | Posted on Feb 8, 2017 at 01:55 UTC as 30th comment | 4 replies
In reply to:

pkcpga: Awesome technology, can't wait to see a larger version. This would definitely help Sony phone sales, only issue is how much memory would your phone need for even just a few short films at 1000 fps and how long can it shoot continuously for. Can't wait to see the final product.

Sort-of negative wait time. Look at the last two RX100. Not exactly the same tech, but very similar, with 1000FPS 1080 (well, it records as 1080, but is actually from fewer pixels) thanks to a 1" sensor backed with DRAM.

Link | Posted on Feb 8, 2017 at 01:39 UTC
On article Leica Summilux-SL 50mm F1.4 sample gallery (363 comments in total)
In reply to:

mxx: Just so everybody knows: I've got nothing to say about this lens.

There, I've said it.

Meh.

Link | Posted on Feb 6, 2017 at 16:38 UTC

Very nice. A tad bizarre that on the right side of the article title it says:

"Buy on Amazon.com From $2,796.95"

What was that selling? ;-)
I don't think this was really about the camera they used.

Link | Posted on Feb 4, 2017 at 16:15 UTC as 49th comment

I had a pair of these with FC-E8 converters tethered to a laptop on a simple autonomous robotic platform to capture fisheye images that that were sent to a cluster supercomputer to stitch and present an interactive 360-degree pan/zoom display on a video wall. That was in our 1999 research exhibit at the IEEE/ACM Supercomputing conference. After that, one became the tethered camera in the ceiling of my machine room... and it's still there, but not running since our last major cluster build party. A shot from it is at http://aggregate.org/SENSORS/108AMARK/bignow.jpg . Last count was over 3,000,000 exposures and still ok.

Link | Posted on Feb 2, 2017 at 22:26 UTC as 46th comment
On article Sony SLT a99 II Review (1481 comments in total)
In reply to:

ProfHankD: An absolutely top-of-the-line camera that still gets little respect in the comments here and is getting reviewed a year after release....

BTW, the comments about lenses are simply incorrect -- if you actually go count how many AF lenses (new+old, but mostly old) are available in A mount it is literally MORE than are native for any other mount (e.g., based on actual counts from places like KEH and eBay). Basically, Minolta was the first with a "modern" AF SLR so everybody made lenses for it -- and they're inordinately cheap now because A mount is no longer as popular. Take a look at Dyxum for a well-maintained list of many of the A-mount lenses with reviews.

HFLM: the 5D IV problem might be triggered by the adapter (?), but the camera locking and SAYING the battery should be removed and reinserted is a demented response which is very clearly how it was designed to behave when it detected some sort of lens error. As for the A-mount lenses, old ones don't have a focus motor -- it's in the body, like Nikon also did -- so it's plenty fast, but noisy (i.e., not ok for video). As for 42MP, almost no lenses with designs older than 3 years were literally designed to handle that -- and certainly not most Canon nor Nikon lenses. That's fine. Being at or beyond a Nyquist sampling of lens resolution is largely a good thing.

Link | Posted on Feb 2, 2017 at 15:53 UTC
Total: 1187, showing: 1 – 20
« First‹ Previous12345Next ›Last »