PIX 2015
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: 725, showing: 21 – 40
« First‹ Previous12345Next ›Last »
On Fujifilm announces X-T1 IR for infrared photography article (203 comments in total)
In reply to:

EcoR1: UV- spectrum of light is from ~ 200 to 400nm. This camera should not be advertised as a true UV-camera in the key features as the sensor can see only a very small portion of UV-light. Of course the biggest drawback for UV- photography are lenses. Most glass-materials used in lens elements don't have transmission below ~350nm. So even if the sensor would record lower than 380nm, it wouldn't do any good without specific and very expensive UV-lenses.

As I said below, 1000nm is ok for NIR, but some sensors go as far as 1200nm. There is still the lens issue for that too, because it is at best unclear that Fuji has designed any of their lenses to keep everything in focus across the full 380-1000nm range. If it is just supposed to be NIR, then shouldn't they have built-in a visible-light blocking filter? For that matter, their lenses might not even keep a consistent focal plane within the 700-1000nm NIR band that they do cover -- many lenses don't.

Anyway, this is picking nits. Fuji's just allowing you to get a more raw version of their camera without a 3rd-party mod... that's all.

Direct link | Posted on Aug 3, 2015 at 18:01 UTC
On Fujifilm announces X-T1 IR for infrared photography article (203 comments in total)

380-1000nm isn't particularly wide spectrum. There are cameras that go as far into NIR as 1200nm, and 380nm is just barely into NUV/UVA. My guess is that they didn't worry about the 380nm limit because most glass barely passes anything beyond 400nm, so they'd be lens-limited before that anyway. I don't see any special wide-spectrum lenses being announced here....

Anyway, still cool that they offer it as a factory version rather than having to get one modified by a 3rd party....

Direct link | Posted on Aug 3, 2015 at 15:35 UTC as 24th comment
In reply to:

The Squire: If the idea is to replace system RAM with this tech it better be *much* more than 1,000 more resilient than current NAND, because it'll be hit a lot lot more frequently than if it were just used as an alternative to SSD.

That said, in cameras it will work well, as buffer memory is written to infrequently (compared to the main RAM of a PC). The distinction between buffer and card will be removed (assuming someone standardizes a very fast interface to this memory tech).

Zdman: one word: CACHE (and/or explicitly-managed local memory, somewhat like we have in GPUs). Performance on things like image processing would be terrible without caches big enough to hold what you're working on, and I'd expect this new memory tech to be paired with processors having cache sized in GBs. Writes that never make it past cache don't count.

Generally tweaking system software and hardware design to minimize writes to XPoint memory sounds like a great research opportunity for folks like me.... ;-)

Direct link | Posted on Aug 3, 2015 at 15:21 UTC
On Green Dragon, Hobbiton in the Brewpubs challenge (2 comments in total)

Thanks. In fairness, it's a cooler pub than most. ;-)
I could have brought out a bit more detail in the window light, and cut the noise a bit, but I was lazy and this is basically the camera JPEG rather than from the raw (it was shot raw+JPEG).

Direct link | Posted on Aug 3, 2015 at 15:05 UTC as 1st comment
In reply to:

ProfHankD: I applaud Canon for doing something new, and this is a good answer for how to innovate when your fab is outdated. Unfortunately, $30K for 60p 1080 is really not cutting it. The resolution is not where you want it for cinematic releases, and the impressively high ISO is only being half utilized when they don't have any framerates above 60/s (probably limited by Canon's sensor chip architecture).

Well, speaking as a researcher, I care about $30K. ;-)
Too pricey for what it is....

Direct link | Posted on Jul 30, 2015 at 21:40 UTC

I applaud Canon for doing something new, and this is a good answer for how to innovate when your fab is outdated. Unfortunately, $30K for 60p 1080 is really not cutting it. The resolution is not where you want it for cinematic releases, and the impressively high ISO is only being half utilized when they don't have any framerates above 60/s (probably limited by Canon's sensor chip architecture).

Direct link | Posted on Jul 30, 2015 at 13:37 UTC as 68th comment | 5 replies
In reply to:

The Squire: If the idea is to replace system RAM with this tech it better be *much* more than 1,000 more resilient than current NAND, because it'll be hit a lot lot more frequently than if it were just used as an alternative to SSD.

That said, in cameras it will work well, as buffer memory is written to infrequently (compared to the main RAM of a PC). The distinction between buffer and card will be removed (assuming someone standardizes a very fast interface to this memory tech).

DVT80111, The Squire: Actually, 10M write cycles/cell probably is enough to replace DRAM. There are two tricks. One is that the vast majority of "memory" accesses are actually satisfied by on-chip SRAM caches, so most memory writes never actually get to DRAM. The second trick is wear leveling: basically, if you have an object whose value is frequently rewritten, you essentially keep moving the object's physical address so the write count on any one cell doesn't saturate until all cells do. This type of management is already done in SSDs, and is why they're usable despite the 10K write lifespan, but it will take a little new compiler/OS/processor architecture tech to do this with main memory.

My quick little back-of-the-envelope computation of the expected lifespan, given the above tricks, would put first failure around 5-10 years for a typical high-end PC using this instead of DRAM... and ECC (as used in SSDs) could extend that. DRAM actually gets (soft) errors way more often....

Direct link | Posted on Jul 30, 2015 at 12:14 UTC
In reply to:

The Squire: If the idea is to replace system RAM with this tech it better be *much* more than 1,000 more resilient than current NAND, because it'll be hit a lot lot more frequently than if it were just used as an alternative to SSD.

That said, in cameras it will work well, as buffer memory is written to infrequently (compared to the main RAM of a PC). The distinction between buffer and card will be removed (assuming someone standardizes a very fast interface to this memory tech).

Just to put some numbers: current flash memory can only survive about 10K write cycles per cell, so this must be around 10M. The same kind of "wear leveling" (deliberate cycling of which blocks are written) should make 10M enough for the life of most consumer devices. And yes, conventional systems use DRAM raw buffers and write things to flash which is ~1500x slower, whereas this sounds to be within a factor of 2 of DRAM speed, so you could skip DRAM and treat the non-volatile storage for a file directly as a memory-mapped raw buffer.

This is potentially a very big deal. I teach the intro computer architecture course at the University of Kentucky and the implications of this new tech might be the biggest change I've had to make to the course content in a decade.

Direct link | Posted on Jul 30, 2015 at 03:53 UTC

Well, there goes the memory heirarchy.... ;-)

This should be a no-brainer win for cameras, enabling things like raw 4K video.

Direct link | Posted on Jul 29, 2015 at 16:52 UTC as 13th comment
On Readers' Showcase: Luis dos Santos article (53 comments in total)

Nice use of HDR! Probably shocking to many people (if not so surprising for me), that the dynamic range of lighting in buildings requires multi-shot HDR processing with a naturally about-as-high-DR-as-you-can-get camera like the A7R.

Direct link | Posted on Jul 26, 2015 at 12:13 UTC as 28th comment | 3 replies
In reply to:

Mssimo: I wonder if they could make speed booster for Medium format lenses onto full frame (sony a7) cameras.

noirdesir: I think you're confused here -- nobody was talking about putting a DSLR behind a focal reducer (although such a reducer could possibly be made for P6 and other medium format lenses). Both u4/3 and Sony E are mirrorless. The lens being used for both u4/3 and E focal reducers is generally a 135-format SLR lens, although you could use a medium format lens on a glassless adapter. The u4/3 and E flange distances are short enough to fit a focal reducer fairly easily -- although making a u4/3 reducer is harder, because the smaller sensor wants a stronger reducer AND u4/3 has a 2mm longer flange distance. Putting a u4/3 reducer on an E body is simply sticking a 2mm glassless adapter behind the u4/3 reducer.

Sensor size only impacts vignetting with lenses from a different mount. The key is flange distance for the target body being shorter than that of the lens by enough to fit the reducer, otherwise you can't focus to infinity without adding a (weak) teleconverter.

Direct link | Posted on Jul 26, 2015 at 02:16 UTC

I stopped using these shockingly-expensive designer camera bag/holder thingies long ago. Am I the only one here who regularly uses things like a little backpack soft cooler ($7 from Walmart) to carry serious camera gear?

To be precise, on longish trips I'll usually nest a smaller $5 bag in that densely holding 3-4 lenses, batteries, memory cards, cleaning supplies, passport, etc. Then I can easily fit my A7II, NEX-7, and a waterproof compact camera on top of that nested bag in the backpack, all loose but padded by half the lenses being in padded pouches in addition to the bag padding. That $7 outer bag is even very waterproof; it kept everything dry visiting Niagara Falls Cave of the Winds. Heck, it even has external mesh pockets for a water bottle!

My cheap solution is way more functional. I can flexibly subset my gear and use the smaller bag alone for side trips. I also can nest that whole little backpack in my regular backpack that holds my laptop, chargers, etc.

Direct link | Posted on Jul 25, 2015 at 04:14 UTC as 33rd comment | 2 replies
In reply to:

Mssimo: I wonder if they could make speed booster for Medium format lenses onto full frame (sony a7) cameras.

Actually, the existing Speed Booster (or other focal reducers) can be useful on FF Sonys using 135-format SLR lenses too -- they will not cover the full frame that way, but can allow things like a square crop with the same view angle as the lens would get uncropped on FF without the focal reducer. You can also get an uncropped >180 degree circular fisheye view that way using one of the Samyang 8mm f/3.5 CSII rectangular APS-C fisheye lenses. It's a little special purpose, but it's a nice trick to have available....

In fact, there's nothing preventing you from mounting the u4/3 SB on a Sony E-mount body via a u4/3-to-E adapter. Not sure, but that might give similar benefits on an E APS-C body to using the APS-C SB on an E FF body. Such u4/3-to-E adapters are of course thin & unchipped, so no AF doing that....

Direct link | Posted on Jul 25, 2015 at 01:50 UTC
In reply to:

Naveed Akhtar: This really look promising to me ..
It opens many fronts .. a small concealable camera for better wild-life photography.
For drones flights .. and other risky places where you can't reach or stay too long.
Then developer's sdk can extend its capabilities beyond current limitations.
Price is also very good for this little and capable camera.
well all I can say is good luck with the project .. and I will definately buy this one over oly air or any other action or mobile-adon camera .. soon!

Yup; Sony opened this door, and nobody has really had the right product walk through yet. What you really want is this sort of thing with CHDK-like programmability and some wired user I/O interfaces. People haven't quite gotten the idea yet that these cameras should be intelligently taking the photos under their own control, not just be dumb 802.11 remotes.

PS: There is a Q&A that says an SDK will be available, but no details, so not very high hopes....

Direct link | Posted on Jul 21, 2015 at 03:51 UTC
On Pentax K-3 II added to studio test scene comparison article (177 comments in total)

My 24MP reference body is the A6000. This Pentax can beat it for preserving detail at high ISOs using pixel shift, but actually looks a touch less sharp and slightly noisier... the up side is that it has a lot less moire, which is precisely what I'd expect. I would like to see the Sony A7II and A7RII provide a similar feature using their IBIS, I'd also like to see this not be a tripod only thing....

Direct link | Posted on Jul 16, 2015 at 21:45 UTC as 50th comment

Ok, another expensive f/0.95. Ho hum. That's easy to do cheaper with a focal reducer. Except I don't know of a heck of a lot of fast, let alone ultrafast, lenses that focal reduce to 10mm, so this is something new. Probably a big deal for video folks....

Direct link | Posted on Jul 15, 2015 at 16:49 UTC as 5th comment
On Sony Cyber-shot DSC-RX100 IV First Impressions Review preview (1551 comments in total)

I want to hear more about 960FPS. Sony has built a number of sensors (e.g., used in some Casios) that did such high framerates, but at very low resolution -- tens of thousands of pixels/frame. This sensor sounds capable of much higher resolution at 960FPS, if not for very long per sequence. I'm very interested in how long you can record at 960FPS with how short a gap between capturing such high-framerate clips.

Direct link | Posted on Jul 14, 2015 at 20:22 UTC as 269th comment
On Phase One 645DF+ with IQ250 field test article (146 comments in total)

Damien: The multi-colored out-of-focus point spread function (OOF PSF, aka bokeh pattern for a point) in the last shot on page 3 is truly odd. I've measured OOF PSFs for 140+ lenses and never seen anything like that. Does that really happen for the OOF PSF with various point sources? Do you have any idea what caused it? If it isn't an artifact from the point sources having lens properties (many artificial lights have lenses), my guess would be this is a sensor reflection perhaps involving the Bayer filter colors being bounced between a couple of lens elements or even the aperture blades themselves....

PS: Leaf shutters inherently tend to (slightly) improve bokeh because the open/close essentially means the exposure covers a range of apertures and not just the taking aperture, slightly apodizing along the lines of the STF emulation mode in the Minolta Maxxum 7.

Direct link | Posted on Jul 11, 2015 at 13:03 UTC as 12th comment | 2 replies
In reply to:

Birdkai: Please guys, stop ignoring the facts Canon really weaker in this regard. The message of Canon lagging behind really needs to deliver to them by having a weak sales and bigger sales in Sony sensor. Wait! Don't get angry, what i mean is let Canon really get back into R&D of sensor, not just self contain with high market share, we need competition in this. Don't you want better tech in sensor? If Sony won in this, they will be stagnant as well, keep on pushing slight improvement in every variance. Samsung did great in the NX1, now A7RII is out, lets just keep rolling new tech that benefits us all.

Sony is winning the big hybrid chip sensor fab battle... against the very aggressive Samsung. Canon isn't a contender in this contest, still analog with huge feature sizes.

Incidentally, Samsung is the default winner in most chip fab contests these days, with Intel being the other dominant high-end player; Intel wins on the parts they make, but Samsung seems to be making a wider range of parts for more companies. Former fab leaders IBM and AMD have sort-of left the fab business, which is unfortunately an ongoing trend.

Direct link | Posted on Jul 10, 2015 at 11:35 UTC
In reply to:

photo perzon: So the Nikon is much better?

The DR of the Canon 5DS is not terrible, and in flat enough lighting it's fine; simple JPEGs only encode about 9 EV. However, the DR is virtually identical to the 11.2 EV of the Sony A100, which means Canon is behind Sony sensor tech by about a decade. Getting close to 13 EV DR matters a lot for general shooting, because that roughly matches instantaneous human vision.

In sum, DR is not a battle Canon fab technology can win. They'll need to leverage the fact that their very conservative fab should produce large sensors at a much lower unit cost. I bet Canon could get a 50MP FF EOS-M-style body out for under $1000... that would be interesting, especially running Magic Lantern (ML). Keep in mind that ML's dual-ISO exposure hack, if it can be applied to that sensor, could get you 13+ EV in a single shot.

Direct link | Posted on Jul 10, 2015 at 04:34 UTC
Total: 725, showing: 21 – 40
« First‹ Previous12345Next ›Last »