• Please use real names.

    Greetings to all who have registered to OPF and those guests taking a look around. Please use real names. Registrations with fictitious names will not be processed. REAL NAMES ONLY will be processed

    Firstname Lastname

    Register

    We are a courteous and supportive community. No need to hide behind an alia. If you have a genuine need for privacy/secrecy then let me know!
  • Welcome to the new site. Here's a thread about the update where you can post your feedback, ask questions or spot those nasty bugs!

Is that in hot dogs per furlong?

Doug Kerr

Well-known member
An article on the site PetaPixel.com:

http://petapixel.com/2015/06/16/canon-5ds-and-5ds-r-initial-resolution-tests/

(perhaps somehow associated with Lens Rentals) reports on initial tests of the measured resolution of the new Canon EOS 5DS and 5DS R dSLR cameras.

Sadly, the author (Roger Cicala) does not bother to state the units of the various results. And it's not obvious from the reported values. Does anyone know what unit might be involved here? For the moment, we will assume that it is in hot dogs per furlong.

Here is the table of results using a Zeiss 85mm f/1.4 Otus lens, consider to itself be a very "high-resolution" lens:

chart3.jpg
From PetaPixel.com​

Now, for comparison, I reviewed an article on the site of Amateur Photographer (UK):

http://www.amateurphotographer.co.uk/reviews/dslrs/canon-eos-5ds-r-review/7

which reviews resolution tests on the EOS 5DS R only. Testing with a Canon EF 100mm f/2.8 Macro USM at f/4, they find that the camera gives a resolution of about 5600 lines/picture height. (This is the "TV line" convention, in which the white spaces between the black lines on the target count as "lines".)

That is quite stunning, given that the sensor in that camera has 5792 sensels in the vertical direction. That result would imply a Kell factor of 96.7%. But that's how it looks on the test chart image they showed.

Now, how do I reconcile that with Cicala's finding of 1570 hot dogs per furlong? I don't even try.

Best regards,

Doug
 

Doug Kerr

Well-known member
Hi, Asher,

Doug,

See how long per mm or lose per inch fits in!

I don't understand.

Do you mean "lines per mm" and "lines per inch"?

I'll try that.

OK. 5600 lines per picture height, for this sensor (24 mm high), would be:

233 lines/mm (or 117 cycles per mm)

5926 lines/inch (2963 cycles per inch)

Thanks.

Best regards,

Doug
 

Doug Kerr

Well-known member
Now, putting aside the matter of units (something I was never able to get my engineering school professors to do when grading my laboratory reports) it is interesting to note, in Cicala's article:

• With The Zeiss 85mm f/1.4 Otus lens aboard (no mention of aperture used), the resolution of the 5DS R (at the center of the frame) was 5.7% greater than that of the 5DS. The resultion of the 5DS here was 1570 hot dogs/furlong.

• With the Canon EF 300 f/2.8 IS II lens aboard the resolution of the 5DS R (at the center of the frame) was 12.4 % greater than that of the 5DS. The resultion of the 5DS here was 1375 hot dogs/furlong.

• With the Canon EF 50mm f/1.4 lens aboard (no mention of aperture used), the resolution of the 5DS R (at the center of the frame) was 1.3% greater than that of the 5DS. The resultion of the 5DS here was 790 hot dogs/furlong.

One wouild be tempted to think that the better the resolution afforded with a certain lens, the more we would see a difference in the resolution of the 5DS R (with no anti-aliasing low pass filter) over the 5DS. But it dies not seem to fully come out that way.

There could of course be a number of technical explanations for that.

Recall that the price paid for the greater resolution of the 5DS R is possible aliasing on subjects with high-frequency components (fine detail), which (especially in the case of recurrent fine detail) may result in moiré patterns in the delivered image. The aliasing would typically be greater for the "green" "channel" (owing to its lower Nyquist frequency), perhaps leading to some false color artifacts.

Best regards,

Doug
 

Asher Kelman

OPF Owner/Editor-in-Chief
Doug,

You need to define the new terms "resultion" or does it just get used instead of resolution in the units, hot dog per furlong", LOL!

Still, can you try to see if there's a fit to either lp/mm or else lp/inch with the unit less data provided in that article?

Asher
 

Doug Kerr

Well-known member
Hi, Asher,

Doug,

You need to define the new terms "resultion" or does it just get used instead of resolution in the units, hot dog per furlong", LOL!

Ops! And I guess this somehow got legitimatized in my spell checking dictionary!

Still, can you try to see if there's a fit to either lp/mm or else lp/inch with the unit less data provided in that article?

Well, at 5600 lines per picture height, for this sensor (24 mm high), those would be:

117 line pairs per mm

2963 line pairs per inch

Best regards,

Doug
 

Asher Kelman

OPF Owner/Editor-in-Chief
Hi, Asher,



Ops! And I guess this somehow got legitimatized in my spell checking dictionary!



Well, at 5600 lines per picture height, for this sensor (24 mm high), those would be:

117 line pairs per mm

2963 line pairs per inch

Best regards,

Doug

From this, one must infer, it seems, that the article uses lp/inch, (unless truly he means hotdogs/furlong, LOL).

The consequence of this would be, that at the center, under the best conditions, only ~~ 50% of the resolution capability of the new sensor is being used.

Can you comment on that?

Asher
 

Doug Kerr

Well-known member
Hi, Asher,

From this, one must infer, it seems, that the article uses lp/inch, (unless truly he means hotdogs/furlong, LOL).

I do not conclude that at all. If we accept that assumption, then (as you point out below) the findings are not credible.

The consequence of this would be, that at the center, under the best conditions, only ~~ 50% of the resolution capability of the new sensor is being used.

Can you comment on that?

Yes. I do not believe that those results are stated in line pairs/inch.

Usually at best we can expect to get image resolution (in lines/picture height) on the order of perhaps 85%, maybe 90%, of the vertical pixel dimensions of the image (the sensel dimensions of the sensor).

Best regards,

Doug
 

Jerome Marot

Well-known member
The article was first published by LensRentals. If you go to the LensRental site, you will find out other lens tests, for example this one. It appears that the units are Line Pairs per picture height and that the measurement correspond to 50% MTF. They use Imatest for their measurements.

Imatest have a page explaining their idea of 50% MTF and what the importance of "standardised sharpening" here.

Probably the higher figures you cited correspond to Lines per picture height (not Line Pairs), measured at extinction (something like 10% MTF).
 

Doug Kerr

Well-known member
Hi, Jerome,

The article was first published by LensRentals. If you go to the LensRental site, you will find out other lens tests, for example this one. It appears that the units are Line Pairs per picture height and that the measurement correspond to 50% MTF.

Thank you.

They use Imatest for their measurements.

Imatest have a page explaining their idea of 50% MTF and what the importance of "standardised sharpening" here.

Thank you. I think in fact I had seen that a while ago. I will review it.

Probably the higher figures you cited correspond to Lines per picture height (not Line Pairs), measured at extinction (something like 10% MTF).

Yes. It was in lines/picture height based on visual examination of a "trumpet" pattern with the reported resolution at what I would call extinction.

Thanks for getting this straightened out.

Best regards,

Doug
 

Doug Kerr

Well-known member
So, with the way spiffy Zeiss Otus lens aboard the 5DS R, the resolution at the center (defined on an MTF50 basis) is reported to be 1666 cycles/PH (often spoken of as 1666 LP/PH).

The Nyquist frequency of the sensor (on a sensel basis, matching the pixel dimensions of the largest deliverable image) is about 2900 cycles/PH.

Interesting.

And we have a new unit synonymous with cycles/PH: HotD/fur.

Best regards,

Doug
 

Doug Kerr

Well-known member
The Imatest system used for the determination of resolution utilizes what is called the "slanted edge" approach. Those interested in the principle of this approach may be interested in the article "Determining MTF with a Slant Edge Target", available here:

http://dougkerr.net/Pumpkin/articles/MTF_Slant_Edge.pdf

The article includes an extensive explanation of the concept of the modulation transfer function (MTF) itself.

Best regards,

Doug
 

Asher Kelman

OPF Owner/Editor-in-Chief
So, with the way spiffy Zeiss Otus lens aboard the 5DS R, the resolution at the center (defined on an MTF50 basis) is reported to be 1666 cycles/PH (often spoken of as 1666 LP/PH).

The Nyquist frequency of the sensor (on a sensel basis, matching the pixel dimensions of the largest deliverable image) is about 2900 cycles/PH.

To what extent could this mean that the apparently best class lens is only able to use about 57% of the resolution of the 5Ds R sensor?

or put another way, would the same detail be resolved with a lesser sensor and if so, how many "5DS R quality" pixels would be needed?

or we just look at a picture of something with fine detail and vote?

Asher
 

Doug Kerr

Well-known member
Hi, Asher,

To what extent could this mean that the apparently best class lens is only able to use about 57% of the resolution of the 5Ds R sensor?

We cannot think that the potentially-attainable "resolution" of a sensor (of itself) is its Nyquist frequency.

For, one thing, the fact that this is a CFA ("Bayer") sensor militates against so doing.

Even without that, Kell showed that the best resolution we can actually get in a practical situation is perhaps 85% of the Nyquist frequency. A simplistic understanding of the phenomenon behind this is to consider a test pattern of alternate black and white lines with the line width equal to the pixel pitch of the sensor. For the "ideal" position of the test pattern, we will see the pattern perfectly captured. But move the pattern by (at the sensor) half the line width, and the image will show just uniform gray. So what is the resolution of this camera?

So Kell's work suggested that, considering that this exact situation is unlikely to occur in real scenes, on the average, over the various portions of the various scenes, we can expect the resolution to be perhaps at best 85% of the Nyquist frequency.

Now I don't know that if with a "perfect" lens (one whose resolution is so great that it does not in any way affect the resolution of the lens+sensor combination) an MTF 50 resolution of 57% of the Nyquist frequency is the best we can expect of a CFA sensor. Others here may know.

would the same detail be resolved with a lesser sensor. . .

Probably not (unless the result in the case of interest was substantially limited by the resolution of the Zeiss Otus lens, and I have not looked into that).

It's like auto engines. Does displacement count? Yes. ("Inches count"). Does compression ratio count? Yes. Does "mixture" optimization (today, injection optimization) count? Yes. Does valve timing optimization count? Yes.

Best regards,

Doug
 

Asher Kelman

OPF Owner/Editor-in-Chief
"Militates"? That's a very nice work.

Do you mean mitigates, Doug, or my vocal is impaired. I'll look it up! :)

Well, now, I am vocabulary challenged!

MILITATE (of a fact or circumstance) be a powerful or conclusive factor in preventing.
"these fundamental differences will militate against the two communities coming together"
synonyms: tend to prevent, work against, hinder, discourage, prejudice, be detrimental to
"his resentment of others in the company militates against his own chances for advancement"​

Now that's cleared up! Thanks for the explanation.

Asher
 

Doug Kerr

Well-known member
The impact of the use of a CFA array sensor (rather than a true "trichromatic pixel" sensor) on resolution potential is not easy to predict.

But we can get the point with a very naïve and simplistic model. That model says that to determine the color of a "spot" on the scene requires a cluster of four sensels (one so-called R, one so-called B, and two so-called G).

Under that admittedly-simplistic view, the potential resolution of the sensor would be half its "sensel resolution" (half its sensel Nyquist frequency).

Now of course the way we demosaic the output of a CFA sensor does not follow that simplistic view, but is much more "clever", so we do better than that.

But that outlook should give us some appreciation of what we are fighting here.

Best regards,

Doug
 

Doug Kerr

Well-known member
I had hoped to be able to estimate the resolution potential of the Zeiss 85 mm f/1.4 Otus lens. but I have been frustrated by the inability to find the applicable MTF curves in the form needed for that.

All the MTF curves (including those on the manufacturer's data sheet) are in the form that shows the modulation transfer "coefficient" as a function of distance from the center of the image for various values of spatial frequency.

But of course what we need to assess "resolution" is the form that shows the modulation transfer "coefficient" as a function of spatial frequency, perhaps for various distances from the center of the image.

Now we might get lucky, if one of the spacial frequency curves in the form of the curves that are available happens to show a modulation transfer "coefficient" of just half that shown by the curve for a pretty low frequency. But no such luck.

In the manufacturer's curves, the highest spatial frequency for which a curve is shown is 40 cy/mm. If we consider the tests at f/4, that has a modulation transfer coefficient of about 90% that for a spatial frequency of 10 cy/mm.

Now can we infer from that at what spatial frequency would the modulation transfer coefficient be 50% of what it was at, say, 3 cy/mm (the criterion for the "MTF50" definition of resolution)?

No.

So, screw it.

Best regards,

Doug
 

Jerome Marot

Well-known member
I had hoped to be able to estimate the resolution potential of the Zeiss 85 mm f/1.4 Otus lens.


And, in effect, you just found out that there is no such think as "resolution potential". Just a function showing how much the signal is attenuated at a given resolution.
 

Asher Kelman

OPF Owner/Editor-in-Chief
And, in effect, you just found out that there is no such think as "resolution potential". Just a function showing how much the signal is attenuated at a given resolution.


Essentially, Jerome, you're like the engineers Psychotherapist, saying it's OK that we cannot express such things.

In practice, it seems like it's back to real pictures.

But, is there still room, from the MTF data, for venturing to say whether the lenses are "sufficient" or lacking to take full advantage of such sensel-packed camera for certain types of photography, where resolving detail is a significant factor, say at 10cm from the eye?

Asher
 

Jerome Marot

Well-known member
Essentially, Jerome, you're like the engineers Psychotherapist, saying it's OK that we cannot express such things.

Not at all. I did not say that.

In practice, it seems like it's back to real pictures.

Cameras and their lenses are meant to take pictures. There is no "back to real pictures" because there never was any "away from real pictures". I just came here to answer a question about units.

But, is there still room, from the MTF data, for venturing to say whether the lenses are "sufficient" or lacking to take full advantage of such sensel-packed camera for certain types of photography, where resolving detail is a significant factor, say at 10cm from the eye?

There is always room to take the time to understand the characteristics of the tools we have to take pictures. Tools to understand them might include abstract notions like MTF curves, yes. MTF characterises sharpness just as much as it characterises a lens that is so unsharp that, in a recent thread, you wondered whether it was malfunctioning.
 

Asher Kelman

OPF Owner/Editor-in-Chief
Essentially, Jerome, you're like the engineers Psychotherapist, saying it's OK that we cannot express such things.

Jerome Marot said:
Not at all. I did not say that.

Of course! My remark was meant as humor. I was comforted by your remarks about our own limits in using MTF data.

In practice, it seems like it's back to real pictures.



Jerome Marot said:
Cameras and their lenses are meant to take pictures. There is no "back to real pictures" because there never was any "away from real pictures". I just came here to answer a question about units.

I simply referred to our ultimate need to simply inspect real images, not just MTF results. I had nothing in my mind of "Real v. "Unreal". All pictures, of course, are real!


Asher Kelman said:
But, is there still room, from the MTF data, for venturing to say whether the lenses are "sufficient" or lacking to take full advantage of such sensel-packed camera for certain types of photography, where resolving detail is a significant factor, say at 10cm from the eye?

Jerome Marot said:
There is always room to take the time to understand the characteristics of the tools we have to take pictures. Tools to understand them might include abstract notions like MTF curves, yes. MTF characterises sharpness just as much as it characterises a lens that is so unsharp that, in a recent thread, you wondered whether it was malfunctioning.

I did? What lens was that?


Asher
 

Asher Kelman

OPF Owner/Editor-in-Chief
I thought about your thread about the Visionar 60mm 1.8

Haha! ?

That was functioning pretty fairly for a lens handheld in front of the camera! I have hopes that the MTF of that lens might be better a lot of my other lenses! That lens should be superb when I have it mounted! I just need to find out how far away I can get objects in focus.


Asher
 

Doug Kerr

Well-known member
Hi, Fahim,

Furlong? I thought it was forlorn :)

Do we still use chains.

The chain is of course 1/10 of a furlong.

In the US there is still some use of the unit "chain" in agriculture and, curiously enough, in denominating the rate of spread of wildfires (in chains/hour).

And of course the length of a UK cricket pitch (bey\tween the stumps) is one chain.

I think the unit is no longer used in the US in land surveying.

Best regards,

Doug
 

StuartRae

New member
My grandfather was the local magistrate, and was the keeper of a standard chain which was used to resolve boundary disputes.

Regards,

Stuart
 
Top