The GetDPI Photography Forum

Great to see you here. Join our insightful photographic forum today and start tapping into a huge wealth of photographic knowledge. Completing our simple registration process will allow you to gain access to exclusive content, add your own topics and posts, share your work and connect with other members through your own private inbox! And don’t forget to say hi!

Why did you choose your back?

T

thsinar

Guest
You're welcome, David.

Please keep in mind, that the resolution given is far more in the centre, about 120 lp/mm, as per opposition to the resolution at the edge of the image circle given in the table.

The table can be downloaded from:

www.sinarcameras.com

Best regards,
Thierry

Thierry, excellent table. I have not seen this one. Thank you!

David
 

David Klepacki

New member
60 MPix in a 4.5 X 6 format would result in about the same lp/mm resolution capability of the current M8 sensor. This would create a lens limited platform. Unfortunately, the physics of it all would reduce the signal to noise ratio as well just due to photon shot noise, so some of the MF-ness would go away.
-bob
Bob, I don't mean to pick on you tonight, but you are wrong again. The 60MP camera would not have the same resolution as the M8. It would actually have much more, except in the realm of macro dimensions.

The Leica M8 has a 10MP sensor in a 3:2 configuration resulting in a pixel image of 3936 x 2630.

A hypothetical 60MP MFDB sensor in a 4:3 configuration results in a pixel image of roughly 8944 x 6708, but to compare apples with apples, I will make it the same 3:2 configuration as the M8, or 9486 x 6324.

Now, let's say that the British Museum calls you up and asks you to do some important copy work for them. It involves photographing some artifacts of size 10cm x 10cm. For simplicity, let's just look at the longest dimension and a single image capture. And, to compare apples with apples, we must assume for each camera that a lens and shooting distance are used so as to provide the same field of view and depth of field.

If you took your M8 to do the job, you would be taking 10cm of the information and distribute that over its available pixels of 3936, giving you a resolution of 3936 / 100mm = 39.36 p/mm = 19.68 lp/mm.

If you took your 60MP MFDB to do the job, you would be taking that same 10cm of information and distribute that over the much larger number of pixels, namely 9486. This would result in a resolution of 9486 / 100mm = 94.86 p/mm = 47.43 lp/mm, or more than twice the resolution of the M8.

This MFDB will always provide higher resolution images, up to the limit of the lens. Only above that limit (which can be up to 250 lp/mm for some MF lenses), will the M8 begin to yield higher resolution. However, this will only occur at the macro scale, never at infinity.
 

Bob

Administrator
Staff member
Bob, I don't mean to pick on you tonight, but you are wrong again. The 60MP camera would not have the same resolution as the M8. It would actually have much more, except in the realm of macro dimensions.

The Leica M8 has a 10MP sensor in a 3:2 configuration resulting in a pixel image of 3936 x 2630.

A hypothetical 60MP MFDB sensor in a 4:3 configuration results in a pixel image of roughly 8944 x 6708, but to compare apples with apples, I will make it the same 3:2 configuration as the M8, or 9486 x 6324.

Now, let's say that the British Museum calls you up and asks you to do some important copy work for them. It involves photographing some artifacts of size 10cm x 10cm. For simplicity, let's just look at the longest dimension and a single image capture. And, to compare apples with apples, we must assume for each camera that a lens and shooting distance are used so as to provide the same field of view and depth of field.

If you took your M8 to do the job, you would be taking 10cm of the information and distribute that over its available pixels of 3936, giving you a resolution of 3936 / 100mm = 39.36 p/mm = 19.68 lp/mm.

If you took your 60MP MFDB to do the job, you would be taking that same 10cm of information and distribute that over the much larger number of pixels, namely 9486. This would result in a resolution of 9486 / 100mm = 94.86 p/mm = 47.43 lp/mm, or more than twice the resolution of the M8.

This MFDB will always provide higher resolution images, up to the limit of the lens. Only above that limit (which can be up to 250 lp/mm for some MF lenses), will the M8 begin to yield higher resolution. However, this will only occur at the macro scale, never at infinity.
David,
I think we are talking about two different things entirely,
I was using as a measure of resolution, the on-sensor resolution stated in line pairs per mm, not the total image size nor the total line pairs resolved by the sensor. A lens limited sensor is a sensor that is capable of resolution greater than what lenses can deliver.
I completely agree that MF has more capability to capture a more detailed image when measured across the total sensor size. That is one reason I bought a P45+.
so a table follows- Data taken from product data sheets and posted previously on a different thread:

The columns are:Camera; sensor dimension in mm; pixels max; LP resolvable; LP/mm on-sensor;resolution factor compared to an M8; Maximum Print dimension at 6lp/mm

M8__ 27.0mm 3916 1958 72.5 1.00 326
D3__ 36.0mm 4256 2128 59.1 0.82 354
P20+ 36.9mm 4080 2040 55.3 0.76 340
P25+ 48.9mm 5436 2718 55.6 0.77 453
P30+ 44.2mm 6496 3248 73.5 1.01 541
P45+ 48.9mm 7216 3608 73.8 1.02 601
X60_ 48.9mm 8944 4472 91.5 1.26 745

the X60 is the hypothetical 60MP sensor I was referring to in my earlier post. I chose arbitrarily to make it a 645 1.1 crop sensor for comparison purposes to the P45+
What I was comparing is the fourth column which is the sensor's maximum capability to "test" a lens. This is a theoretical limit based on the Shannon-Nyquist sampling theorem, your actual results will be worse than this.
I don't know of a manufacturer that publishes MTF charts at 90 lines per mm, but if I take a fairly good lens such as the Leica 50mm Summicron, which publishes data at only 40 lines per mm, I can estimate that at 90 lines per mm and f/5.6 that its contrast might be in the center only roughly 25%-50% which is the limit I was referring to. This is true only at the point of best focus, and it drops off away from that point, Of course, any lens would be diffraction limited to about f/9 in a 60 MP 645 sensor.

So I think this data speaks for itself, do you disagree?
thanks
-bob
 

David Klepacki

New member
...
So I think this data speaks for itself, do you disagree?
thanks
-bob
Yes, I disagree. The camera data concerning the sensor size and number of pixels is the only accurate data in the table. The other derived numbers concerning resolution and print size are not correct.

The first reason is the one I already told you. These sensors are not monochromatic sensors, nor are they foveon-type sensors where all three chromaticities are captured at every pixel location. Let's take the M8 as an example. The M8 has only 1968 pixels in the horizontal direction in any one color, NOT 3936. So, the Nyquist limit is essentially half of what is reported in your table.

Second, no matter how generous you are in estimating the resolving capability of the sensor, there is plenty of MF glass that still out resolves it. As an example, take a look at the MTF data from Zeiss for the Hasselblad 40 CFE IF lens, found here , where Zeiss explicitly says that this lens is capable of 200 lp/mm. Also, Kornelius Fleischer has measured many of the Hasselblad V lenses, mostly reported at photo.net, and I do not recall ever seeing anything less than about 160 lp/mm.

The bottom line is that the MFDB sensors are not out resolving decent MF glass, and will not out resolve this glass even with 60MP.

[As an aside, you seem to be interested in obtaining the smallest resolvable feature in an image, yet you do not specify the field of view, magnification and scale of consideration. Without these parameters, I don't know what kind of valid conclusions you will be able to make. Certainly, you can change the magnification with lenses of longer focal lengths. In more extreme cases, you can use more powerful telescopes or microscopes to resolve the features that you are looking for, depending on your desired scale.]
 

Jack

Sr. Administrator
Staff member
Yes, I disagree.
~snip~ The M8 has only 1968 pixels in the horizontal direction in any one color, NOT 3936.
David, I understand what you are saying, but think there is a flaw in your base assumption. I would like to point out that a conventional line-pair is black and white. Black contains no color information, and white contains ALL colors ----- as such, black won't excite any pixel regardless of color while white will in fact excite them all... As for generating them in the PROPER color, that is where the Nyquist limit comes in, so all in all, I think Bob's comments are on point...

Moreover, I have done direct resolution testing of actual test targets with B&W line pairs, and have always found the best lenses hit about 70% of maximum possible sensor resolution, or about Nyquist limits. AND I calculated the maximum theoretical resolution just as Bob did. So at least in that regard, I have some direct empirical support for Bob's points...

Cheers,
 

David Klepacki

New member
David, I understand what you are saying, but think there is a flaw in your base assumption. I would like to point out that a conventional line-pair is black and white. Black contains no color information, and white contains ALL colors ----- as such, black won't excite any pixel regardless of color while white will in fact excite them all... As for generating them in the PROPER color, that is where the Nyquist limit comes in, so all in all, I think Bob's comments are on point...

Moreover, I have done direct resolution testing of actual test targets with B&W line pairs, and have always found the best lenses hit about 70% of maximum possible sensor resolution, or about Nyquist limits. AND I calculated the maximum theoretical resolution just as Bob did. So at least in that regard, I have some direct empirical support for Bob's points...

Cheers,
Jack, there is no flaw in my argument. I happen to be a professional physicist, earning my PhD quite a long time ago, so I know what I am talking about.

First, black and white line pairs are already software fabrications of the captured image. As I said, the sensor only gives you 33% of the image information, none of which is black or white, and the rest of the information must be estimated. The software algorithms to interpolate the missing information have become extremely sophisticated in estimating this information, but it is just that, an estimation.

So, you are not really measuring the Nyquist limit of the sensor, but rather the degree to which the raw conversion is capable of predicting the missing information. As such, you cannot conclude that a larger sensor, like the hypothetical 60MP one proposed here, will be lens limited.

Furthermore, the Nyquist limit is not based on black-white or on any specific color. It is strictly a function of the sampling frequency of the wavelengths it is measuring. So, for the M8, it is physically impossible to obtain resolutions that exceed 36.44 lp/mm in either red or blue, although it is possible to exceed this number in the diagonal direction in green (which is roughly 72 lp/mm). There are simply not enough red or blue photosites to give you more red-blue information than this. Basically, the post-capture interpolation processing provides an "artificial boost" in resolution via software estimation. This is the reason why you are able to measure a higher lp/mm in your own black-white tests. The much higher frequency green information is mixed with the red and blue information in order to generate the black-white content. I guess it would be analogous to how sharpening can artificially improve MTF measurements.

The best way to prove this to yourself is to work with test charts that are not black-white but rather red-blue, as it is much harder for the software to estimate the missing information in this case.
 

Jack

Sr. Administrator
Staff member
WHITE excites EACH color site be it R,G or B to some degree, therefore it renders a signal on the chip. Now, we can argue how accurately that information gets processed, but cannot argue the fact that a single pixel of each color can be excited to some degree by the same source. AND the fact a single site gets excited is enough to give it a part in the role of net output resolution...

Cheers,
 

David Klepacki

New member
Well, if you are only concerned with creating images having full spectrum white detail (or only shades of green), then I guess you are good....but once your images hold more color, your lp/mm will be limited by the red and blue channels as I have tried to describe here.

BTW Jack, I already explained how all R, G, B sites are excited to produce white. It does not change the sampling rates of the red or blue channels though. It is still half the frequency of the green chanel sampling rate.
 
Last edited:

Jack

Sr. Administrator
Staff member
Hmmm.... my test target (not an ISO 12233) holds black and white line pairs along with Red, Blue and Yellow line pairs, arranged in vertical horizontal and tangential orientations. In all the testing I've done, I almost always have at least one lens that will hit the Nyquist limit for the sensor in question, and usually several will, at least centrally on all colors. Moreover, often times yellow or blue are the strongest rendering, showing slightly more resolution than black or red, and sometimes the tangential show a bit better than the vertical or horizontal. Regardless, using your argument, the best I should be able to hit is around half Nyquist with the colored pairs, yet all the same, I frequently hit Nyquist with all colors? So I suspect either you are misunderstanding us, or are possibly misinformed, or Bob, myself and Nyquist are all wrong...
 
D

DougDolde

Guest
I have been looking for one of those yellow stream icons to express my thoughts reading this dribble but can't find one.
 

Guy Mancuso

Administrator, Instructor
Hey Doug after reading all that I am just happy to be a dumb ass enough not to understand any of it.:ROTFL::ROTFL::ROTFL::ROTFL::ROTFL:

I'm in trouble now. I swear it is the heat getting to me
 

Guy Mancuso

Administrator, Instructor
It was pretty funny to see , reason I mentioned it and of course i don't want any members getting duped into this stuff. Okay back to the regular program.
 

David Klepacki

New member
Hmmm.... my test target (not an ISO 12233) holds black and white line pairs along with Red, Blue and Yellow line pairs, arranged in vertical horizontal and tangential orientations. In all the testing I've done, I almost always have at least one lens that will hit the Nyquist limit for the sensor in question, and usually several will, at least centrally on all colors. Moreover, often times yellow or blue are the strongest rendering, showing slightly more resolution than black or red, and sometimes the tangential show a bit better than the vertical or horizontal. Regardless, using your argument, the best I should be able to hit is around half Nyquist with the colored pairs, yet all the same, I frequently hit Nyquist with all colors? So I suspect either you are misunderstanding us, or are possibly misinformed, or Bob, myself and Nyquist are all wrong...
Jack, it is certainly possible to see test charts hit the Nyquist limit. This just means that the raw conversion process is extremely good. I can show you test charts where the resolution has even exceeded the Nyquist limit, which should not be physically possible, yet it can happen if the estimation of the missing information is clever enough.

If your test chart is absolutely pure red on a pure blue background, with no green content whatsoever, I would like to know how you expect to obtain a higher Nyquist limit. Unfortunately, I know of no test chart pure enough to eliminate any trace of green, so there will always be some high frequency green content to better interpolate the red and blue.
 

Dale Allyn

New member
Guy Mancuso said:
BTW folks this is a scam . No way you can put a P45 on a H3/39 and still be a H3/39 . Actually pretty lame. Look at the images of it
http://cgi.ebay.com/Hasselblad-H3D-3...QQcmdZViewItem
You're right. Hass back in one photo, P45 in the next! I reported it :)

Alarming that people are bidding on such a scam.


If you read the e-mails on the listing he attempts to clarify. It says it's bundled with two backs and he admits that he screwed up by listing them together.

Here's part of it: "There were several questions concerning the P45. The back is from my old H1, which was stolen from my car last year, and I bought h3d. I bundled up the back for sale with h3d, but on afterthought, it wasn't the best idea, as some people want to buy the back only. Unfortunately, I don't know how to resolve it, as some people already bid on the combo, and it would be unfair to change the auction, now. Sorry. "

Not sayin' I know, just reading what's there.
 
Top