Mike Davis

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: 44, showing: 1 – 20
« First‹ Previous123Next ›Last »
On CP+ 2015: Canon shows off prototype 120MP CMOS sensor article (255 comments in total)
In reply to:

mike earussi: With a pixel size of 2.2um, diffraction will start to destroy the resolution after f2.8, so I really don't see this having any practical application for regular photographers. Nor do I think any lens currently on the market can shoot at this level of resolution at f2.8. So this is a technical achievement, not a practical one, unless it's considered for it's PR value or for bragging rights.

I'd be far more impressed by Canon increasing the DR of its sensors, which would have practical value, instead of just their MP count.

Well said, despite the lack of specifying the enlargement factor and desired print resolution at which f/2.8 would begin to be an issue.

Direct link | Posted on Feb 15, 2015 at 16:47 UTC
On CP+ 2015: Canon shows off prototype 120MP CMOS sensor article (255 comments in total)
In reply to:

John Crawley: When is the industry going to learn that more MP isn't the answer.

I agree with James123 and mosc. Your posts are like a breath of fresh air.

Direct link | Posted on Feb 15, 2015 at 16:44 UTC
On CP+ 2015: Canon shows off prototype 120MP CMOS sensor article (255 comments in total)
In reply to:

Frank_BR: Many good lenses produce details in the center of the field which can only be revealed by a sensor resolution from 200 to 500 MP. Therefore, the increase in sensor resolution is most welcome. Many people who use the argument of diffraction against increasing sensor resolution forget that the impact of diffraction is gradual, and much of the falling of the response can be compensated via digital processing.

8 lp/mm is generally accepted as the highest resolution any adult with healthy vision can appreciate at a viewing distance of 10 inches. So, even if you desire a print resolution of 8 lp/mm (in a non-resampled 576 dpi print), there's no point in having a 120 MP sensor if you intend to make prints smaller than 15.9 x 23.1 inches - the size you'd get using all 120 MP to secure 8 lp/mm - and remember, even this size print, at this resolution, demands that you avoid stopping down below f/4.6 - thanks to diffraction making all those pixels useless if you do so.

If you intend to make lower than 5 lp/mm resolution prints at that enlargement factor with a 120 MP sensor, then again, you don't need 120 MP.

In short, the only way to actually take advantage of all those pixels on so small a sensor is to forget about using most of the f-Numbers offered by your lenses - thanks to diffraction. Never mind the signal-to-noise ratios suffered with such a tiny pixel pitch.

Direct link | Posted on Feb 15, 2015 at 16:24 UTC
On CP+ 2015: Canon shows off prototype 120MP CMOS sensor article (255 comments in total)
In reply to:

Frank_BR: Many good lenses produce details in the center of the field which can only be revealed by a sensor resolution from 200 to 500 MP. Therefore, the increase in sensor resolution is most welcome. Many people who use the argument of diffraction against increasing sensor resolution forget that the impact of diffraction is gradual, and much of the falling of the response can be compensated via digital processing.

At f/9, with this 32.1x enlargement factor, you might as well have used a 60MP sensor, because diffraction will reduce your print resolution to 2.5 lp/mm, the equivalent of 180 dpi after AA and RGBG losses.

At f/18, you'd have been fine with a 30MP sensor, because diffraction will reduce your print resolution to 1.25 lp/mm, the equivalent of 90 dpi after AA and RGBG losses.

Is the print too large in my example, above? If you intend to make smaller prints with a 120 MP sensor, then you don't need 120 MP, unless you desire more than the 5 lp/mm (360 dpi) resolution at the print, used in my example - when most people are making large prints at lower resolutions, because they either don't have the pixels to warrant larger prints and/or they assume no one is going to be scrutinizing them from a distance of only 10 inches.

Direct link | Posted on Feb 15, 2015 at 16:24 UTC
On CP+ 2015: Canon shows off prototype 120MP CMOS sensor article (255 comments in total)
In reply to:

Frank_BR: Many good lenses produce details in the center of the field which can only be revealed by a sensor resolution from 200 to 500 MP. Therefore, the increase in sensor resolution is most welcome. Many people who use the argument of diffraction against increasing sensor resolution forget that the impact of diffraction is gradual, and much of the falling of the response can be compensated via digital processing.

Yes, the effect of diffraction comes on gradually, but the f-Number at which diffraction will just begin to inhibit a desired print resolution in lp/mm, at an anticipated enlargement factor can be calculated as follows:

Max. f-Number = 1 / desired print resolution / anticipated enlargement factor / 0.00135383

Running the numbers for a desired print resolution of 5 lp/mm (equivalent to 360 dpi after AA and RGBG losses), we get

Max f-Number = 1 / 5 / 32.1 = 4.6

Thus, with a 32.1x enlargement factor (for a 25.5 x 36.9-inch 360 dpi print), any attempt to use f-Numbers larger than f/4.6 will reduce the print resolution, thanks to the gradual onset of diffraction.

No amount of post-processing can restore genuine subject detail that was lost due to diffraction at the time of exposure. Acuity (edge sharpness) can be improved, but lost resolution (actual subject detail) cannot be created from nothing.

Direct link | Posted on Feb 15, 2015 at 16:24 UTC
On CP+ 2015: Canon shows off prototype 120MP CMOS sensor article (255 comments in total)
In reply to:

Frank_BR: Many good lenses produce details in the center of the field which can only be revealed by a sensor resolution from 200 to 500 MP. Therefore, the increase in sensor resolution is most welcome. Many people who use the argument of diffraction against increasing sensor resolution forget that the impact of diffraction is gradual, and much of the falling of the response can be compensated via digital processing.

Despite the gradual onset of diffraction, if anyone tries to use a 120MP capture to make non-resampled 360 dpi prints (equivalent to 5 lp/mm after losses in resolution caused by the AA filter and RGBG algorithm), the resulting print would measure 25.5 x 36.9 inches, while suffering an outrageous enlargement factor of 32.1x from the smaller than full-frame sensor measuring only 20.2 x 29.2mm.

At that enlargement factor, diffraction's Airy disks, for any given f-Number, will be magnified so much in the final print as to force the use of f-Numbers no greater than f/4.6 to deliver a desired print resolution of 5 lp/mm - to actually make use of the resolution promised by the 120 MP sensor.

Direct link | Posted on Feb 15, 2015 at 16:23 UTC
On CP+ 2015: Canon shows off new EOS 5DS and 5DS R article (123 comments in total)
In reply to:

MarkByland: Sort of false representation taking 100 photographs, stitching them together, and presenting them as some thing that would come straight out of camera. Why not do side-by-side 1:1 series taken from a MkIII? Or, wait, a D810? Show us what you've got, not what can be done with major digital processing.

Also, does this 5Ds R come with a free, multiple terabyte, cloud based storage account?

I would nevertheless be irritated at seeing those stitched prints. Show me what the camera itself can do with a single exposure.

Direct link | Posted on Feb 15, 2015 at 15:27 UTC
On Hasselblad unveils pixel-shifting 200MP H5D-200c MS article (230 comments in total)
In reply to:

Mike Davis: For a print resolution goal of 5 lp/mm (unresampled 360 ppi):

In single-shot mode, with 51.3 MP on a 43.8x32.9mm sensor, diffraction would begin to inhibit a desired print resolution of 5 lp/mm (about 360 ppi) in a 23.0 x 17.2-inch print at f/11.1. (Sensor pixel density is 189 pixels/mm. Enlargement factor for that print size is 13.33x.)

In six-shot mode, with 205.2MP on the same 43.8x32.9mm sensor, diffraction would begin to inhibit a desired print resolution of 5 lp/mm (about 360 ppi) in a 46.0 x 34.4-inch print at f/5.5. (Sensor pixel density is 377 pixels/mm. Enlargement factor for that print size is 26.65x.)

For a print resolution goal of only 2.5 lp/mm (unresampled 180 ppi):

In single-shot mode, with 51.3 MP on a 43.8x32.9mm sensor, diffraction would begin to inhibit a desired print resolution of 2.5 lp/mm (about 180 ppi) in a 46.0 x 34.4-inch print at f/11.1. (Sensor pixel density is 189 pixels/mm. Enlargement factor for that print size is 26.65x.)

In six-shot mode, with 205.2MP on the same 43.8x32.9mm sensor, diffraction would begin to inhibit a desired print resolution of 2.5 lp/mm (about 180 ppi) in a 92.0 x 68.8-inch print at f/5.5. (Pixel density is 377 pixels/mm. Enlargement factor for that print size is 53.3x.)

What size print do you hope to make, at what resolution? Do you need this camera's 6-shot mode and are you willing to shoot at apertures wide enough to exploit the higher pixel density without inhibiting your desired print resolution?

Direct link | Posted on Aug 22, 2014 at 20:25 UTC
On Hasselblad unveils pixel-shifting 200MP H5D-200c MS article (230 comments in total)

For a print resolution goal of 5 lp/mm (unresampled 360 ppi):

In single-shot mode, with 51.3 MP on a 43.8x32.9mm sensor, diffraction would begin to inhibit a desired print resolution of 5 lp/mm (about 360 ppi) in a 23.0 x 17.2-inch print at f/11.1. (Sensor pixel density is 189 pixels/mm. Enlargement factor for that print size is 13.33x.)

In six-shot mode, with 205.2MP on the same 43.8x32.9mm sensor, diffraction would begin to inhibit a desired print resolution of 5 lp/mm (about 360 ppi) in a 46.0 x 34.4-inch print at f/5.5. (Sensor pixel density is 377 pixels/mm. Enlargement factor for that print size is 26.65x.)

Direct link | Posted on Aug 22, 2014 at 20:25 UTC as 47th comment | 2 replies

I suspect the label "Diffraction Correction" exaggerates the effectiveness of this feature. No amount of processing can magically recreate actual subject detail that was lost to diffraction as the light passed through the aperture. It might be able to simulate what appears to be genuine subject detail, but it won't be accurate.

For example, assuming that all other variables affecting resolution are up to the task... If diffraction at a given f-Number is just bad enough to prevent you from discerning the date "2014" on the face of a coin lying on a table several meters from a camera equipped with a normal FL lens when viewing at 100%, "Diffraction Correction" isn't going to reconstruct that data from thin air when the data never got past the aperture in the first place.

You can't make a silk purse from a pig's ear.

Direct link | Posted on Jul 23, 2014 at 12:18 UTC as 7th comment | 12 replies
On Stream your photos... via backpack? article (34 comments in total)

He looks deservedly smug...

Direct link | Posted on May 19, 2014 at 13:49 UTC as 15th comment
On Orion DVC210 DLSR Crane Review article (38 comments in total)

For still shots, I'd rather use my Bogen/Manfrotto 3048 and a step ladder to get my camera to a height of 11 feet.

Direct link | Posted on Mar 30, 2014 at 09:15 UTC as 8th comment | 2 replies
On Nikon D4s First Impressions Review preview (1042 comments in total)

ISO 409,600 is eight stops faster than ISO 1600.

Fairground rides: 1/250 sec. @ f/16
Night street scene: 1/200 sec. @ f/16
Interior by candlelight: 1/60 sec. @ f/16
Landscapes by full moon: 1/2 sec. @ f/16

Direct link | Posted on Feb 25, 2014 at 22:55 UTC as 182nd comment | 6 replies
On Cold Hearted Wind in the Bales challenge (4 comments in total)

Trite, but true: I LOVE this!

Direct link | Posted on Nov 8, 2013 at 23:58 UTC as 3rd comment

Sadly, it took 2nd place in this challenge.

Direct link | Posted on Oct 19, 2013 at 12:45 UTC as 1st comment
On Nikon D800 Review preview (18 comments in total)

Regarding the excellent demo of diffraction's impact at various f-stops, on page 25 of this review, where is the photo showing what could have been accomplished using Photoshop ACR sharpening against an f/4 RAW file?

I would very much like to compare sharpening of the f/4 RAW file to sharpening of the f/22 RAW file. Surely, a silk purse made from silk would be more attractive than a silk purse made from a pig's ear.

Link to page 25 of this review: http://www.dpreview.com/reviews/nikon-d800-d800e/25

Mike

Direct link | Posted on Oct 19, 2013 at 12:13 UTC as 11th comment
In reply to:

Humboldt Jim: Can we assume that a 1" sensor system can be stopped down to ƒ16, or even 22 without diffraction problems?

Using the example I've given above, if you're willing to assume that no one will view your prints at distances less than 20 inches (instead of 10 inches), you can double the calculated f-Number, stopping down to f/12.84 - delivering an effective 2.5 lp/mm at 20 inches that will appear every bit as detailed as a 5 lp/mm print at 10 inches.

Please note that there are only two variables in the formula for calculating the f-Number at which diffraction will begin to inhibit a desired print resolution (expressed in lp/mm for a 10-inch viewing distance): Enlargement factor and desired print resolution at a given viewing distance.

Somehow, discussions of resolution (or "sharpness") almost always neglect these critical variables.

Have a look at the equation for calculating the maximum acceptable diameter for a circle of confusion. It includes Enlargement factor and desired print resolution at a given viewing distance.

http://en.wikipedia.org/wiki/Circle_of_confusion

Direct link | Posted on Nov 3, 2012 at 18:28 UTC
In reply to:

Humboldt Jim: Can we assume that a 1" sensor system can be stopped down to ƒ16, or even 22 without diffraction problems?

A desired resolution of 5 lp/mm (which is equivalent to 360dpi, taking into account the 30% loss of resolution imposed by a typical CMOS sensor's Bayer algorithm and AA filter), to support viewing distances as close as 10-inches (25cm), for a 23x enlargement factor (which is required to make an 8x12-inch print from a 1-inch sensor's uncropped capture - an enlargement factor that would require, at 360 dpi, an image resolution of 2880x4320 pixels, or 12.44 MP:

f-Number = 1/ 5 / 23 / 0.00135383 = 6.42

Thus, a one-inch, 12.44 MP sensor cannot deliver more than 5 lp/mm (360dpi) of resolution after enlargement to an 8x12-inch print when using f-Numbers larger than f/6.42.

A one inch sensor is still quite small, compared to a full frame or MF sensor.

Continued below...

Direct link | Posted on Nov 3, 2012 at 18:26 UTC
In reply to:

Humboldt Jim: Can we assume that a 1" sensor system can be stopped down to ƒ16, or even 22 without diffraction problems?

The f-Number at which diffraction will begin to inhibit a desired print resolution (expressed in line pairs per millimeter for a viewing distance of 10 inches), at an anticipated enlargement factor can be calculated as follows:

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

A 1-inch sensor would have dimensions 13.2 x 8.8mm.

The possible f-Numbers at which diffraction would inhibit a desired print resolution at an anticipated viewing distance are endless, but here is an example combination:

Continued below...

Direct link | Posted on Nov 3, 2012 at 18:25 UTC

If this catches on, Facebook will have to buy a lot more hard drives.

Direct link | Posted on Sep 24, 2012 at 17:23 UTC as 33rd comment
Total: 44, showing: 1 – 20
« First‹ Previous123Next ›Last »