[biz-engineer]

Finally, making some prints and looking close, I found so easy to remember rules of thumb for enlargement and sensor resolution.
I rule out the popular pixel count and 300 PPI rule for print sizes, because one can have a 100Mpixel smartphone camera and the A1 sized print from it looking completely fuzzy (looking close).


1) Max enlargement to retain sharp detail in prints even looking close (assuming shutter speed is fast, and low ISO):
-> Size of the sensor in mm , expressed in inches for the print size. Very easy.
-> Micro43: sensor size 13.5mm x 18mm => print size 13.5 inches x 18 inches [ 35 cm x 45 cm ~A3 ]
-> Apsc camera: sensor size = 15 mm x 24 mm => max print size 15 inches x 24 inches [ 40 cm x 60 cm , or A2 ]
-> Full frame camera: 24 mm x 36 mm => print size 24 inches x 36 inches [ 60 cm x 90 cm, ~A1 ]
-> MF crop: 44 mm x 33 mm => print size 44 inches x 33 inches [ ~110 cm x 84 cm , ~A0 ]


2) Sensor resolution:
Lens max resolution: 50 lppmm in the center, down to 30 lppmm in corners.
Minimum sampling rate (nyquist) for each primary color on CFA: 50 lppmm x 2 = 100 pixels per mm (100 pixels/mm for R, 100pixels/mm for G, 100 pixels/mm for B)

Pixel count for FF sensor:
-> X direction : 36 x 2 x 100 = 7200.
-> Y direction: 24 x 2 x 100 = 4800.
-> Pixels count on sensor: 34.56 Mpixels

Pentax K1 sensor has the minimum pixel count for lenses that resolve 50lppmm. A higher pixel density collects fast fadding microcontrast information, contributing less and less to image detail as pixel count increases, but taking a lot more sorage space.

[yucatanPentax]

Very interesting notations. This concept seems like it might be somewhat subjective: "retain sharp detail in prints even looking close."


Just as a point of reference, a friend regularly sells images that are more like 6 x 10 feet in size, printed on metal or other materials, all taken with an old Nikon D800 and a couple of old lenses. I suppose I could put my nose up within a couple feet of the image and see that it is "fuzzy" (is it?), but across even a 10' wide room (many are in hallways), they look perfectly fine. In fact, they're stunning.

He's a master of capturing the light just right with the kinds of things people want to see in office buildings, dealerships, government offices, magazine covers, etc, etc, etc. But he hasn't worried about being sharp enough yet at huge sizes.

Given that, maybe there's some sort of "Immense Image Sharpening" that can be done? I honestly don't know. I'm not trying to take his place, just admire his work.

[biz-engineer]

This concept seems like it might be somewhat subjective: "retain sharp detail in prints even looking close."

For the DFA28-105 at 35mm on K1, it's the enlargement baseline at which I can't tell the difference on print between corners and center, even looking close with prescription glasses. I realized that I have to distinguish between a photo print, and a poster. I'd define a photo print as an object that a person can hold in front of her to appreciate the quality, as opposed to a poster supposed to be displayed on panel and viewed at a distance. So my rule of thumb is for a high quality photo print as an object people can hold and look at away and up close.

---------- Post added 23-01-20 at 11:44 ----------

printed on metal or other materials

The print media can create a very different impression of the same photograph, well beyond mere image detail.

[ToddK]

-> Apsc camera: sensor size = 15 mm x 24 mm => max print size 15 inches x 24 inches [ 40 cm x 60 cm , or A2 ]

I think it depends on the subject of the image. I am sitting in my office looking at a couple of 20"x30" aluminum prints taken with my K-3II and from 5 feet away they look great to me. Both are building architecture shots. I am sure that there are some kinds of shots that would not look good at that size.

[BarryE]

There is an argument that printing is in itself an art rather than an formulaic decision. Perhaps there are too many numbers being considered here? Perhaps ...

[stevebrot]

Working with your actual pixel resolution is a good starting place. Converting that to instructions for the printer driver is quite another.

Here are a few thoughts:

• Sensor size is irrelevant
• Optical enlargement to the sensor is irrelevant
• Upsampling is extrapolation. If you can live with the detail loss, go for it.
• There are a few other threads on this site that discuss printing strategy in some detail. I have them bookmarked somewhere and will post the links if I find them.
• The software used for printing counts

FWIW, I print from Lightroom from a virtual copy based on an edited softproof to the Printer/Paper ICC profile. The Canon Pixma Pro-100 is directed to use the designated ICC profile and I usually print at 600dpi (pixels and dots are not directly mappable). I print 13x19 max.

Have fun!


Steve

[biz-engineer]

Upsampling is extrapolation. If you can live with the detail loss, go for it.There are a few other threads on this site that discuss printing strategy in some detail. I have them bookmarked somewhere and will post the links if I find them.The software used for printing counts

Yes, I understand / agree with that.

Sensor size is irrelevant. Optical enlargement to the sensor is irrelevant

About those two first points, I like to know more about the background for coming to such conclusion.

Knowing that most recent camera models have sensors with 5um (or less) pixel pitch, I consider sensor size instead of pixel resolution. I could observe on print that the edges and corners show some blur (from the lens) which is independant from lack of sharp detail by lack of pixels. When corners start to look fuzzy, the middle of the print can still be zoomed in an extra 30% without seeing distracting blur. Just to test if my logic is flawed or not, I consider a digital sensor with infinite resolution, what's left is the optical bandwidth and sensor size. And given what I can see with the K1, tells me that the A7RIV would be of little benefit except for cropping the center area of the image when the optical resolution is maximum.

A better definition (to complete my initial post) would be:
- if the pixel pitch is >= than 5um, consider pixel resolution and PPI for estimating print size.
- if the pixel pitch is < 5um, consider sensor size for estimating print size.

---------- Post added 23-01-20 at 18:40 ----------

I think it depends on the subject of the image.

That's not included in my post, and yes, relative size of the subject in the frame is key. I looked at the gallery prints of David Yarrow (wildlife photographer and Nikon ambassador) who used a D810 and D850 for capturing his photographs , then 1:1 square crops printed at 48", 60" and 70". I asked myself "how can he blow up D810 images that much?" then I understood the "trick": the subject covers as much as 30% to 50% of the frame, whether it shows the face of a lion very close, or elephants (he stands away from the camera that he triggers using a remote control device).

---

Last edited by biz-engineer; 01-23-2020 at 10:42 AM.

[stevebrot]

About those two first points, I like to know more about the background for coming to such conclusion.

I've done a fair amount of printing?

Though sensor size, pixel pitch, and optical enlargement are considerations for capture quality, none are relevant for printing in that the damage or optimization is already done by that point. Capture and PP to your best quality and you will do well on the printing end. Anything more is grossly overthinking the process with little or no dividend to be had in the end.

I print from scanned 35mm, 6x7, and 4x5 negatives, 10Mpx from the K10D, and 24Mpx from the K-3 and get suitable-for-framing results to 13x19 from any file suitable to print.


Steve

[biz-engineer]

I print from scanned 35mm, 6x7, and 4x5 negatives, 10Mpx from the K10D, and 24Mpx from the K-3 and get suitable-for-framing results to 13x19 from any file suitable to print.

I understand, this makes sense.

[JensE]

A better definition (to complete my initial post) would be:- if the pixel pitch is >= than 5um, consider pixel resolution and PPI for estimating print size.- if the pixel pitch is < 5um, consider sensor size for estimating print size.

The rule-of-thumb for sensor size, as you explained it, is based on the assumption that the above described the transition from a sensor-pixel limitation to other limitations, notably finite lens resolution. There have been substantial improvements over the last decade in lenses, at least in what is accessible to photographers. There is nothing preventing e.g. an APS-C sensor with equal pixel as a full-frame sensor to deliver the same print-ability within the limits for diffraction and dynamic range. The ~3um sensors in 4/3 cameras still seem to deliver 'real' resolution with good lenses at their optimum, fairly fast, aperture.

[ne!]

A little trick I use when preparing to print is to view my image in Lightroom at 100% on my 15" MacBook Pro which has a resolution of 220 pixels per inch. If it looks good at this magnification, I can rest assured that a print at 220 DPI will have the same sharpness.

In the case of the K-1, the math is as follows: 7360x4912 at 36x24 inches gives me about 204 DPI.

So basically it'll be just a bit softer than what I see on my screen at 100% magnification. For other print sizes, I usually divide the pixel dimension by the number of inches to determine my printing DPI, so a 30 x 20 inch print is 7360 / 30 = 245 ppi. If I'm happy with my image viewed on-screen at 220 dots per inch, then my print at 245 DPI will be even sharper.

As an aside, I recently made a 20x30 inch print from a 4896 × 3264 image shot with my GR and even at 163 PPI the print turned out great.

[BarryE]

A little trick I use when preparing to print is to view my image in Lightroom at 100% on my 15" MacBook Pro which has a resolution of 220 pixels per inch. If it looks good at this magnification, I can rest assured that a print at 220 DPI will have the same sharpness.

In the case of the K-1, the math is as follows: 7360x4912 at 36x24 inches gives me about 204 DPI.

So basically it'll be just a bit softer than what I see on my screen at 100% magnification. For other print sizes, I usually divide the pixel dimension by the number of inches to determine my printing DPI, so a 30 x 20 inch print is 7360 / 30 = 245 ppi. If I'm happy with my image viewed on-screen at 220 dots per inch, then my print at 245 DPI will be even sharper.

As an aside, I recently made a 20x30 inch print from a 4896 × 3264 image shot with my GR and even at 163 PPI the print turned out great.

But Epson's native dpi is 360 and HP and Canon is 300. Best to set ppi to match. If use different resolutions then this should be taken into account. As I said above printing is an art as well as a science. There's too much grinding through these type of figures when the paper, subject, output sharpening, viewing distance etc etc, all making up a big impact in the final print without the numbers - sensor size becomes less relevant.

Also, the amount of ink being laid down is critical at each point as some papers saturate more easily and thus blur.

[biz-engineer]

There have been substantial improvements over the last decade in lenses, at least in what is accessible to photographers. There is nothing preventing e.g. an APS-C sensor with equal pixel as a full-frame sensor to deliver the same print-ability within the limits for diffraction and dynamic range. The ~3um sensors in 4/3 cameras still seem to deliver 'real' resolution with good lenses at their optimum, fairly fast, aperture.

There have been improvements on lenses, no doubt about it. I compared many MTF charts of both primes and zooms, and I found out that the majority of lenses have 30lppmm microcontrast drops between 20% (best lenses) down to 60% (even good lenses), around the outer quarter of frames. All lenses design trade weight and size for some loss of microcontrast away from the center. 1:1 crop can be printed as large at a non cropped 3:2 image. Only rare and expensive lenses such as Zeiss Otus , Pentax D-FA*50 have almost flat MTF charts.

---------- Post added 24-01-20 at 18:36 ----------

A little trick I use when preparing to print is to view my image in Lightroom at 100% on my 15" MacBook Pro which has a resolution of 220 pixels per inch. If it looks good at this magnification, I can rest assured that a print at 220 DPI will have the same sharpness.

That's a good trick. I did that with my notebook (180 ppi) because display pixel density is higher than the density on my 27" display, and the smeared corners I see on the notebook display look almost the same as on physical print. Zooming in 100% using a 27" 4K monitor would give an even closer simulated look.