Mike Davis

Lives in United States Dallas, TX, United States
Has a website at http://www.accessz.com
Joined on Jun 12, 2002

Comments

Total: 86, showing: 1 – 20
« First‹ Previous12345Next ›Last »
On article Google AI adds detail to low-resolution images (148 comments in total)
In reply to:

Biowizard: I'll believe this when Google provides a web page where I can upload an arbitrary 8*8 pixel photo reduction, and get something back that vaguely resembles my original photo. Until then, it's unproven snake oil.

Brian

@Biowizard Yes, and can they do it without reliance on higher resolution versions of the images already known to Google - the first step of their two-step approach?

Send them a photo of a coin held between someone's thumb and forefinger, rendered at a resolution similar to the samples shown above, then see if Google can recover the coin's year of issue.

Link | Posted on Feb 10, 2017 at 18:08 UTC
In reply to:

Mike Davis: With 8256x6192 pixels on a 43.8x32.9mm sensor, I've calculated a pixel density of 188.4 pixels/mm, which falls below my long-held opinion that, ideally, sensors should never exceed a maximum of 200 pixels/mm, lest the cameras suffer an inadequate range of f-Numbers at which diffraction will support true subject detail resolution of 5 lp/mm (in a non-resampled, uncropped, 360 dpi final print from a CMOS sensor, where the RGBG Bayer algorithm and AA filter will reduce a pixel count-implied resolution by 30%), to support a viewing distance as close as 25cm (9.84 inches).

Thanks for stating that, explicitly. It makes misinterpretation less likely.

And thus, you're saying the same thing I said with this equation:

f -Number = 1 / desired print resolution in lp/mm / anticipated enlargement factor / 0.00135383

Rearranged:

Print resolution in lp/mm = 738.645 / (f-Number * anticipated enlargement factor)

Note the absence of any variable for pixel count or pixel density.

We really are in agreement. It only looks as if we aren't when qualifications are omitted or ignored as we try to use the English language to convey scenarios.

Link | Posted on Jan 26, 2017 at 07:05 UTC
In reply to:

Mike Davis: With 8256x6192 pixels on a 43.8x32.9mm sensor, I've calculated a pixel density of 188.4 pixels/mm, which falls below my long-held opinion that, ideally, sensors should never exceed a maximum of 200 pixels/mm, lest the cameras suffer an inadequate range of f-Numbers at which diffraction will support true subject detail resolution of 5 lp/mm (in a non-resampled, uncropped, 360 dpi final print from a CMOS sensor, where the RGBG Bayer algorithm and AA filter will reduce a pixel count-implied resolution by 30%), to support a viewing distance as close as 25cm (9.84 inches).

But what display size and what viewing distance?

Link | Posted on Jan 26, 2017 at 03:08 UTC
In reply to:

Mike Davis: With 8256x6192 pixels on a 43.8x32.9mm sensor, I've calculated a pixel density of 188.4 pixels/mm, which falls below my long-held opinion that, ideally, sensors should never exceed a maximum of 200 pixels/mm, lest the cameras suffer an inadequate range of f-Numbers at which diffraction will support true subject detail resolution of 5 lp/mm (in a non-resampled, uncropped, 360 dpi final print from a CMOS sensor, where the RGBG Bayer algorithm and AA filter will reduce a pixel count-implied resolution by 30%), to support a viewing distance as close as 25cm (9.84 inches).

You've not stated a print size and viewing distance for your comparison scenario, above. This makes your contention impossible to prove. Really, all you've given us is yet another example of people talking apples and oranges for failure to fully qualify their contentions.

Link | Posted on Jan 25, 2017 at 22:58 UTC
In reply to:

Mike Davis: With 8256x6192 pixels on a 43.8x32.9mm sensor, I've calculated a pixel density of 188.4 pixels/mm, which falls below my long-held opinion that, ideally, sensors should never exceed a maximum of 200 pixels/mm, lest the cameras suffer an inadequate range of f-Numbers at which diffraction will support true subject detail resolution of 5 lp/mm (in a non-resampled, uncropped, 360 dpi final print from a CMOS sensor, where the RGBG Bayer algorithm and AA filter will reduce a pixel count-implied resolution by 30%), to support a viewing distance as close as 25cm (9.84 inches).

Correcting my second to the last comment, above, in the interest of accuracy...

The first paragraph should read as follows:

It's the higher enlargement factor required by small, high-density sensors having the same pixel count as larger sensors that forces the use of *larger* apertures (*smaller* f-Numbers), *to yield* smaller Airy disks at the sensor before magnification, to produce like-sized, like-resolution prints.

Link | Posted on Jan 25, 2017 at 13:18 UTC
In reply to:

Mike Davis: With 8256x6192 pixels on a 43.8x32.9mm sensor, I've calculated a pixel density of 188.4 pixels/mm, which falls below my long-held opinion that, ideally, sensors should never exceed a maximum of 200 pixels/mm, lest the cameras suffer an inadequate range of f-Numbers at which diffraction will support true subject detail resolution of 5 lp/mm (in a non-resampled, uncropped, 360 dpi final print from a CMOS sensor, where the RGBG Bayer algorithm and AA filter will reduce a pixel count-implied resolution by 30%), to support a viewing distance as close as 25cm (9.84 inches).

We're saying the same thing, and now you are qualifying your statements nicely.

Link | Posted on Jan 24, 2017 at 17:29 UTC
In reply to:

Mike Davis: With 8256x6192 pixels on a 43.8x32.9mm sensor, I've calculated a pixel density of 188.4 pixels/mm, which falls below my long-held opinion that, ideally, sensors should never exceed a maximum of 200 pixels/mm, lest the cameras suffer an inadequate range of f-Numbers at which diffraction will support true subject detail resolution of 5 lp/mm (in a non-resampled, uncropped, 360 dpi final print from a CMOS sensor, where the RGBG Bayer algorithm and AA filter will reduce a pixel count-implied resolution by 30%), to support a viewing distance as close as 25cm (9.84 inches).

It's the higher enlargement factor required by small, high-density sensors having the same pixel count as larger sensors that forces the use of smaller apertures (larger f-Numbers), and thus, smaller Airy disks at the sensor before magnification, to produce like-sized, like-resolution prints.

In the end, neglecting any discussion of SNR or other traits, looking only at the impact had on DoF and diffraction, small sensors can do everything the large sensors can do, but they leave us with fewer diffraction-free f-Numbers from which to choose, forcing us to shoot closer to wide open, which yields faster shutter speeds for a given amount of light and ISO setting.

Link | Posted on Jan 24, 2017 at 14:52 UTC
In reply to:

Mike Davis: With 8256x6192 pixels on a 43.8x32.9mm sensor, I've calculated a pixel density of 188.4 pixels/mm, which falls below my long-held opinion that, ideally, sensors should never exceed a maximum of 200 pixels/mm, lest the cameras suffer an inadequate range of f-Numbers at which diffraction will support true subject detail resolution of 5 lp/mm (in a non-resampled, uncropped, 360 dpi final print from a CMOS sensor, where the RGBG Bayer algorithm and AA filter will reduce a pixel count-implied resolution by 30%), to support a viewing distance as close as 25cm (9.84 inches).

With equivalent Pixel Count, DoF, Diffraction, Print Size, and Viewing Distance...

Small, high-density sensors can give us the same DoF and diffraction as larger sensors, but with faster shutter speeds at smaller f-Numbers, and thus, confining the photographer to fewer "diffraction-free" f-Numbers from which to choose, of those offered by any given lens.

Large, low-density sensors having the same pixel count, can give us the same DoF and diffraction as smaller sensors, but with slower shutter speeds at larger f-Numbers, and thus, more "diffraction-free" f-Numbers from which to choose, of those offered by any given lens.

Link | Posted on Jan 24, 2017 at 14:52 UTC
In reply to:

Mike Davis: With 8256x6192 pixels on a 43.8x32.9mm sensor, I've calculated a pixel density of 188.4 pixels/mm, which falls below my long-held opinion that, ideally, sensors should never exceed a maximum of 200 pixels/mm, lest the cameras suffer an inadequate range of f-Numbers at which diffraction will support true subject detail resolution of 5 lp/mm (in a non-resampled, uncropped, 360 dpi final print from a CMOS sensor, where the RGBG Bayer algorithm and AA filter will reduce a pixel count-implied resolution by 30%), to support a viewing distance as close as 25cm (9.84 inches).

"Pixel count has no effect on diffraction. More pixels resolve more detail than fewer pixels stop-for-stop."

This is true when the pixel count is increased while shooting at the same f-Number and maintaining the same viewing distance - without simultaneously increasing the print size.

People with 50 MP sensors will, at the very least, be tempted to make larger prints than they would make with like-sized, 20 MP sensors. If they do make larger prints, the Airy disks produced at any given aperture will suffer the same increase in enlargement factor.

The f-Number at which diffraction will begin to inhibit a desired print resolution, in line pairs per mm, can be calculated as follows:

f = 1 / desired print resolution in lp/mm / anticipated enlargement factor / 0.00135383

Thus, as pixel density increases, all else remaining equal, there's no increase in vulnerability to diffraction. (What Great Bustard said...) Nothing I wrote in my original comment contradicts this.

Link | Posted on Jan 24, 2017 at 14:52 UTC
In reply to:

Mike Davis: With 8256x6192 pixels on a 43.8x32.9mm sensor, I've calculated a pixel density of 188.4 pixels/mm, which falls below my long-held opinion that, ideally, sensors should never exceed a maximum of 200 pixels/mm, lest the cameras suffer an inadequate range of f-Numbers at which diffraction will support true subject detail resolution of 5 lp/mm (in a non-resampled, uncropped, 360 dpi final print from a CMOS sensor, where the RGBG Bayer algorithm and AA filter will reduce a pixel count-implied resolution by 30%), to support a viewing distance as close as 25cm (9.84 inches).

Increase the print size and the viewing distance proportionately (can you enforce this?) and you'll still have to avoid stopping down below f/11.1. Decrease the print size by a factor of two (from 22.93x17.2 to 11.46x8.6 inches), without changing the viewing distance, and you can stop down to f/22.2 for additional DoF, while still delivering subject detail to 5 lp/mm. Reduce your resolution and/or viewing distance requirements and you can make any size print you like.

I don't feel sorry, so much, for Phase One or Hasselblad, but the Pentax 645z has almost the exact same specs, in terms of pixel count, sensor size and, thus, pixel density.

There's much more to choosing a camera than pixel math can dictate and this FujiFilm offering is busting at the seams with nice features. Pentax must be crying.

Link | Posted on Jan 22, 2017 at 15:02 UTC
In reply to:

Mike Davis: With 8256x6192 pixels on a 43.8x32.9mm sensor, I've calculated a pixel density of 188.4 pixels/mm, which falls below my long-held opinion that, ideally, sensors should never exceed a maximum of 200 pixels/mm, lest the cameras suffer an inadequate range of f-Numbers at which diffraction will support true subject detail resolution of 5 lp/mm (in a non-resampled, uncropped, 360 dpi final print from a CMOS sensor, where the RGBG Bayer algorithm and AA filter will reduce a pixel count-implied resolution by 30%), to support a viewing distance as close as 25cm (9.84 inches).

For this 51.1MP sensor, at 188.4 pixels/mm, the maximum permissible 5 lp/mm print size is 22.93 x 17.2 inches (great for a 13x19-inch printer, with room for cropping) AND any f-Number up to f/11.1 can be used without concern for diffraction inhibiting 5 lp/mm subject detail. Double the viewing distance from 25cm to 50cm (can you enforce this?) without changing the print size and you can stop down two additional stops for additional DoF - as far as f/22.2 - as the print resolution falls to 2.5 lp/mm, thanks to diffraction, but looks just as detailed from twice the distance.

Link | Posted on Jan 22, 2017 at 15:01 UTC

With 8256x6192 pixels on a 43.8x32.9mm sensor, I've calculated a pixel density of 188.4 pixels/mm, which falls below my long-held opinion that, ideally, sensors should never exceed a maximum of 200 pixels/mm, lest the cameras suffer an inadequate range of f-Numbers at which diffraction will support true subject detail resolution of 5 lp/mm (in a non-resampled, uncropped, 360 dpi final print from a CMOS sensor, where the RGBG Bayer algorithm and AA filter will reduce a pixel count-implied resolution by 30%), to support a viewing distance as close as 25cm (9.84 inches).

Link | Posted on Jan 22, 2017 at 15:00 UTC as 6th comment | 18 replies

As theaters slowly converted to digital projectors, it was the motion picture industry that prolonged the demise of Kodak film production, well after the majority of roll film still photograghers had rushed into digital cameras.

Interestingly, we can now thank the interest of analog-loving cinematograghers for Ektachrome's resurrection - first, in 35mm motion picture stock, ahead of still photography roll or sheet films.

http://motion.kodak.com/motion/customers/productions/default.htm

Link | Posted on Jan 6, 2017 at 15:06 UTC as 45th comment
On article Dell's 8K monitor goes on sale in March for $5000 (215 comments in total)
In reply to:

Dragonrider: 8K at 32" is just plain silly. I have a 32" 4k monitor and would be much happier if it were at least 40 in. It is a great monitor, but running at 150% in windows is a waste of pixels and 8k would have to run at 300% to be readable. When are we going to see some decent 40-46 in 4k monitors with ips panels and full color gamut?

Sounds logical to me. Might you therefore recommend an 80-inch diagonal for 8K, at the same viewing distance used with a 40-inch diagonal for 4K?

Link | Posted on Jan 6, 2017 at 09:45 UTC
On article Dell's 8K monitor goes on sale in March for $5000 (215 comments in total)
In reply to:

EskeRahn: It does sound a bit like overkill on a 32" monitor, unless you use it really close...

Most people can just distinguish details down to around 1/12000 of the viewing distance. At my age it is more like 1/7000. I have met a single eagled eyed guy with a 1/25000 limit though, I'm sure he would love it.

I have made some test-sheets where you can test your eyes in seconds, that can be found here: http://eskerahn.dk/wordpress/?p=32

Yes! There are lots of people who are already seated too far away from their 4k displays. Almost no one is going to sit closely enough to actually resolve the pixels of an 8k display.

My calculations, per THX recommendations, show that for the following resolutions in a 32-in. diagonal, you can sit no farther than the distances stated if you want to actually enjoy all of the pixels. This assumes you can resolve an angular resolution of 1.0 arc-minute.

For 1080 vertical in a 32-inch, sit no farther than 4.16 ft. (1.27m)
For 2160 (4k) vertical in a 32-inch, sit no farther than 2.08 ft. (0.63m)
For 4320 (8k) vertical in a 32-inch, sit no farther than 1.04 ft. (0.32m)

For 4k displays, multiply the diagonal by 0.78 to get the maximum viewing distance.
For 8k displays, multiply the diagonal by 0.39 to get the maximum viewing distance.

If you buy an 8k display and insist on sitting at the max. distance for a 4k display (or farther still), you might as well have purchased a 4k display!

Link | Posted on Jan 6, 2017 at 00:40 UTC
In reply to:

AstroStan: "color distracts you from the image"

And goosed-up color, even more so. I can't help but notice that nearly every photo in the recent awards is hyper-saturated. But I suppose that's what grabs the competition for eyeballs, even if it's only a 1-3 second glance. This aesthetic runs rampant in pretty pic astro-imaging.

Exactly. I think it starts with super-saturated television. Our culture is jaded to color.

Link | Posted on Dec 31, 2016 at 17:05 UTC
In reply to:

Mike Davis: Canon really has been making improvements with each revision of this series. For this latest $2199 edition, look at the DxOMark Sharpness > Field Map for a FL of 16mm FL at f/8:

https://www.dxomark.com/Lenses/Canon/Canon-EF-16-35mm-F28L-III-USM-mounted-on-Canon-EOS-5DS-R---Measurements__1009

Then open a new tab and compare the Sharpness > Field Map for the currently $1399 (Amazon) mark ii version at 16mm and f/8:

https://www.dxomark.com/Lenses/Canon/EF16-35mm-F2.8L-II-USM-mounted-on-Canon-EOS-5DS-R---Measurements__1009

Open yet another tab to compare the Sharpness > Field Map for the currently priced $1749 (Amazon), but discontinued, original version, at 16mm and f/8:

https://www.dxomark.com/Lenses/Canon/Canon-EF-16-35mm-F28L-USM-mounted-on-Canon-EOS-5DS-R---Measurements__1009

https://cdn.dxomark.com/dakdata/measures/CanonEOS5DSR/Result/DakResult/Canon_EF_16_35mm_F28L_III_USM/MTF_v2/MTF_ACUTANCE_16mm_f8.png

https://cdn.dxomark.com/dakdata/measures/CanonEOS5DSR/Result/DakResult/Canon_EF16_35mm_F28L_II_USM/MTF_v2/MTF_ACUTANCE_16mm_f8.png

https://cdn.dxomark.com/dakdata/measures/CanonEOS5DSR/Result/DakResult/Canon_EF_16_35mm_F28L_USM/MTF_v2/MTF_ACUTANCE_16mm_f8.png

https://cdn.dxomark.com/dakdata/measures/SonyA6000/Result/DakResult/Sony_E_10_18mm_F4/MTF_v2/MTF_ACUTANCE_10mm_f8.png

Link | Posted on Dec 30, 2016 at 13:26 UTC
In reply to:

Mike Davis: Canon really has been making improvements with each revision of this series. For this latest $2199 edition, look at the DxOMark Sharpness > Field Map for a FL of 16mm FL at f/8:

https://www.dxomark.com/Lenses/Canon/Canon-EF-16-35mm-F28L-III-USM-mounted-on-Canon-EOS-5DS-R---Measurements__1009

Then open a new tab and compare the Sharpness > Field Map for the currently $1399 (Amazon) mark ii version at 16mm and f/8:

https://www.dxomark.com/Lenses/Canon/EF16-35mm-F2.8L-II-USM-mounted-on-Canon-EOS-5DS-R---Measurements__1009

Open yet another tab to compare the Sharpness > Field Map for the currently priced $1749 (Amazon), but discontinued, original version, at 16mm and f/8:

https://www.dxomark.com/Lenses/Canon/Canon-EF-16-35mm-F28L-USM-mounted-on-Canon-EOS-5DS-R---Measurements__1009

There's more to a lens than DxOMark Sharpness Field Map measurements, but just for grins, open yet another tab to compare these three Canon full-frame lenses to the currently $748 (Amazon) Sony 10-18mm f/4 (for e mount APS-C bodies), at the equivalent 10mm FL and f/8 (which yields more DoF, but no visible diffraction at any given print size and viewing distance that's reasonable for the pixel count) than a 16mm full-frame lens at f/8):

https://www.dxomark.com/Lenses/Sony/Sony-E-10-18mm-F4-mounted-on-Sony-A6000---Measurements__942

This beats the mark i and mark ii versions of the Canon 16-35mm by a country mile (on a 24MP body). But again, there's more to a lens than DxOMark Sharpness Field Map measurements.

Link | Posted on Dec 28, 2016 at 18:08 UTC

Canon really has been making improvements with each revision of this series. For this latest $2199 edition, look at the DxOMark Sharpness > Field Map for a FL of 16mm FL at f/8:

https://www.dxomark.com/Lenses/Canon/Canon-EF-16-35mm-F28L-III-USM-mounted-on-Canon-EOS-5DS-R---Measurements__1009

Then open a new tab and compare the Sharpness > Field Map for the currently $1399 (Amazon) mark ii version at 16mm and f/8:

https://www.dxomark.com/Lenses/Canon/EF16-35mm-F2.8L-II-USM-mounted-on-Canon-EOS-5DS-R---Measurements__1009

Open yet another tab to compare the Sharpness > Field Map for the currently priced $1749 (Amazon), but discontinued, original version, at 16mm and f/8:

https://www.dxomark.com/Lenses/Canon/Canon-EF-16-35mm-F28L-USM-mounted-on-Canon-EOS-5DS-R---Measurements__1009

Link | Posted on Dec 28, 2016 at 18:06 UTC as 7th comment | 2 replies
On article 'Ethereal' takes you to Iceland in 4K (81 comments in total)
In reply to:

jonny1976: have seen better and more interesting video from iceland...but my point is what's the point of this?
time-lapse are interesting for their kind animation style...this video lack this and look just plainly a video shot normally at 60 frame per second....so they really made this video from still frame? i use drone e for video and still and the inspire has super stabilization but at this level to mimic a video is a bit strange.
i still don't understand the choice of time-lapse if the output is just a video.

The scale of those landscapes is what you're failing to appreciate. Nothing made by DJI can fly that fast. The topography of the land is so much better revealed to the eye when time lapse has us flying at multiples of the actual drone speed. The video could have been produced by shooting from a Bell Ranger, but it wasn't. That's the magic of this guy's work.

Link | Posted on Dec 16, 2016 at 13:27 UTC
Total: 86, showing: 1 – 20
« First‹ Previous12345Next ›Last »