[georgweb]

Hi everybody,
as a follow-up of a conversation about Ole's field-of-view table (which I have misunderstood in the first place) I would like to ask all of you a simple question and I hope many of you can contribute to give a solid answer:

What is the real horizontal width of your imaging sensor ?

I did a quick test shoot with my K100D and came up with the answer:

22,8mm (common wisdom is 24,0mm)

Obviously it all depends on the real focal length of the lens used. I took the fl number from a Zeiss data sheet, who are known to be exact. Here are my measurements:
Camera: K100D
Lens: Contax Planar 50/1.4
fl: 51,8mm
distance: 3860mm
width: 1700mm

I attach an excel sheet à la Ole's table and a quick calculation method on the right side where you only have to put in distance and width to see the FOV. You might have to change the term WURZEL to SQRT for your excel to run properly. Width and distance can be of any unit as long as they are the same and use decimal commas.

Column B and row 5 can be changed to your liking and the whole table will change as a consequence. I really hope I've done the math right this time !

If we have enough data we might also proclaim reference lenses in order to have verified focal length measurements for them. Once your sensor width is verified, it would be easy to determine any lenses' focal length too.

I would say for the test shots to be valid one would have to take shots with a subject as far away as possible in mid-frame. Measuring the distances is of course the flipside of the coin .-)

Looking forwards to your measurements,
Georg
PS: Another very nice table from Venturi

Attached Files

Horizontal_FOV.zip (6.4 KB, 258 views)

[Wheatfield]

I'm just wondering what relevancy this sort of thing has?

[ytterbium]

I'm just wondering what relevancy this sort of thing has?

Probably very small to none in artistic and creative photography.

Most likely a lot for technical use. E.g. you can compare the exact sensor size with available sensor specifications to know if it's the same sensor.

Combined with exactly known lens parameters you can use camera to do some measurements. Calculate real life distances from images.

This might be especially useful with older cameras like k100d that allow remote capture to computer.
You can use the camera in some sort of automated system. E.g. QC during manufacturing.
It's a pity that newer cameras cant be used for tasks like this. Makes them somewhat not professional - take off the release socket, flash hotshoe and something more, you get a bulky consumer p&s.

[Fogel70]

What is the real horizontal width of your imaging sensor ?

I did a quick test shoot with my K100D and came up with the answer:

22,8mm (common wisdom is 24,0mm)

Obviously it all depends on the real focal length of the lens used. I took the fl number from a Zeiss data sheet, who are known to be exact. Here are my measurements:
Camera: K100D
Lens: Contax Planar 50/1.4
fl: 51,8mm
distance: 3860mm
width: 1700mm

I doubt this calculation is correct. Usually FL of a lens changes with focus, and FL specified on a lens is when focus is set on infinity. So what you have found out is that FL of the lens used is not 51.8mm when focus is set to 3860mm.

---

Last edited by Fogel70; 10-23-2010 at 10:47 AM.

[Lowell Goudge]

The specifications say 23.4 x 15.6 for the k7

Additionally wikipedia says that the crop factor of the K20/k/ is 1.54 where as earlier pentaxes was 1.53

[georgweb]

I'm just wondering what relevancy this sort of thing has?

Depends on where your threshold lies as a tolerable divergence. 2%, 5%, 10% ? The other thing I want to get a grip on is to verify the focal lengths of prime and zoom lenses.

...
Combined with exactly known lens parameters you can use camera to do some measurements. Calculate real life distances from images.
This might be especially useful with older cameras like k100d that allow remote Ycapture to computer. ...

Do you refer to some kind of tethering ?

I doubt this calculation is correct. Usually FL of a lens changes with focus, and FL specified on a lens is when focus is set on infinity. So what you have found out is that FL of the lens used is not 51.8mm when focus is set to 3860mm.

I am afraid you are correct. I am thinking of coming up with a very simple way to use some landmarks, google maps or similar to get correct infinity values, especially for long tele-lenses. Edit: Did the exact same test as before with the lens set to the known inf. mark. All else being the same, the width now is 1720mm (out of focus but still very well discernable, @f/1.4). This gives for a horizontal image taking width of the sensor of 23,08mm.

The specifications say 23.4 x 15.6 for the k7
Additionally wikipedia says that the crop factor of the K20/k/ is 1.54 where as earlier pentaxes was 1.53

Allright let's assume we have (image taking) sensor widths of 22,8mm to 23,5mm, somewhere in between there. As I said before I am really curious to find out the (image taking part of) sensor width and as a second step the real focal lengths of lenses. I also like the idea of putting into perspective the razzle-dazzle (including an abundance of so-called tests and reviews) vs. the real thing.

I hope we can come up with the real thing soon .-)
Best, Georg

[Lowell Goudge]

I'm just wondering what relevancy this sort of thing has?

Me too

Even after giving the dimension it is not accepted

[Lowell Goudge]

Depends on where your threshold lies as a tolerable divergence. 2%, 5%, 10% ? The other thing I want to get a grip on is to verify the focal lengths of prime and zoom lenses.



Do you refer to some kind of tethering ?

accept the published value and use it to scale / verify the focal length

I am afraid you are correct. I am thinking of coming up with a very simple way to use some landmarks, google maps or similar to get correct infinity values, especially for long tele-lenses. Edit: Did the exact same test as before with the lens set to the known inf. mark. All else being the same, the width now is 1720mm (out of focus but still very well discernable, @f/1.4). This gives for a horizontal image taking width of the sensor of 23,08mm.



Allright let's assume we have (image taking) sensor widths of 22,8mm to 23,5mm, somewhere in between there. As I said before I am really curious to find out the (image taking part of) sensor width and as a second step the real focal lengths of lenses. I also like the idea of putting into perspective the razzle-dazzle (including an abundance of so-called tests and reviews) vs. the real thing.

I hope we can come up with the real thing soon .-)
Best, Georg

[georgweb]

Me too
Even after giving the dimension it is not accepted.
...accept the published value

No, never :-)

...and use it to scale / verify the focal length

If you have a verified lens you can start with, that would be correct. If not, all calculations would be off.


My information so far is that my K100D should have a published (imaging area - ) sensor width of 23.6mm (K7 23.4mm) whereas my measurements come up with a rounded 23.1mm. That is a 2% difference.

I would go from here and put my measured value in the attached excel sheet in order to get real, horizontal FOV's and real focal lengths. It is using the ARCTAN function (not the crop factoring method) so I am sure that the values will fullfill my high requirement standards .-)

Thanks everybody for participating,
Georg

[Fogel70]

I would trust the dimension given on sensor size more than focal length. But as all values are only given with one decimal, the 2% difference could easily be found in rounded off values.
It could be FL of 51.76mm rounded off to 51.8mm, but a part of it could also be rounded off values of sensor size.

You need values with at least 2 decimals for calculations like this.

[georgweb]

You need values with at least 2 decimals for calculations like this.

Probably, and it looks like the published sensor sizes are in fact very close to what I have found. Apart from that there is always a small part of the sensor being used for calibration and not for imaging (afaik like a thin dark frame around the edges).

[ytterbium]

Btw, if i wanted to help, what kind of lens should i use?

P.S.
I think the published sensor size might not match the imaging area.
That means the size includes masked pixels for black level adjustments and other periphery. It might even include the whole chip size (with pins and base), instead of just the visible area.

[Fogel70]

Btw, if i wanted to help, what kind of lens should i use?

P.S.
I think the published sensor size might not match the imaging area.
That means the size includes masked pixels for black level adjustments and other periphery. It might even include the whole chip size (with pins and base), instead of just the visible area.

If checking data sheet of sensors it seems as if it's only the size of used pixel area that the manufactures specify.

Sony (IMX021) 12.47MP sensor use a sensor diagonal of 28.47mm (crop factor 1.52). With pixel size of 5.49x5.49um and number of active pixels 4312x2880 (12.41MP) makes maximum active sensor area of 23.67x15.81mm.
http://www.sony.net/Products/SC-HP/cx_news/vol50/pdf/imx021.pdf

I think this is the same base sensor Pentax use in K-x and K-r with active pixel area of 4288x2848, which makes a sensor area used of 23.54x15.63mm.

[georgweb]

Btw, if i wanted to help, what kind of lens should i use?

I would say a fast fifty with the focal length known from a trusted source. I just did that and trusted the factory-published focal length (usually not a good idea) as a reference.

What I did is put the lens on full open and infinity and room's length pointing onto a shelf and tried to exactly fill the frame with it. I've added a scale protuding on one side cause it was easier to align the frame only on one side and add a bit through the scale on the other. I was able to read the scale albeit out of focus. I am wondering wether stopping down would change anything in this situation ?

The distance is measured in a centrical line from the object to the camera. Measure up to the bayonet flange and add 45.46mm.

I remember some (old) lens tests verifying the focal length by testing it and publishing it. An updatable list of tested focal lengths would be nice.

[kh1234567890]

For the K-7 it looks suspiciously like 5um per resulting pixel, i.e. 4672 x 0.005 = 23.36mm