My Experiments => Attn. Asher

Hello Asher,

Here is an example of some experiments I did on HFD focus experiments as a "tutorial" for myself.
The tape measure was at HFD/2 and the red tip of the valve was at the HFD. This is one of a series of experiments testing the idead of focus at HFD (Kerr and Prevost) and focus at infinity (Merklinger).

And here is my polarisation experiment (the weather that day was really bad !)



Regards,

Jeremy.

Hi, Jeremy,

Jeremy Waller said:

Hello Asher,

Here is an example of some experiments I did on HFD focus experiments as a "tutorial" for myself.
The tape measure was at HFD/2 and the red tip of the valve was at the HFD. This is one of a series of experiments testing the idead of focus at HFD (Kerr and Prevost) and focus at infinity (Merklinger)

Click to expand...

Glad to hear of your experiments in this area.

Is there any place we can see the results at a pixel size from which we can see what happened?

Thanks.

Best regards,

Doug

Hi, Jeremy,

Jeremy Waller said:

Hello Doug,

Here are 3 pictures:

I knew it - the photos are being resized somewhere.

Click to expand...

Evidently!

I suspect the "/s128/" in the path is a clue that this is a set of resized images (these are 128 px high).

Ah - here they are:


(They can run but they can't hide!)

Now I gotta look at them and figure out what they tell me!

Hard to figger the last one - maybe it's buggered.

Nice work.

Best regards,

Doug

Doug,

Are you imagining the images or you can see them! Except for the first tiny ones, the others do not show up for me! However, you seem to be talking to ghosts that I can't see!

Asher

Here's the last image Doug. When all 3 look good I'll write an explanation.


OK! The image looks right.

This was a series of experiments designed to enable me to see for myself what the effect of focusing at the HFD and infinity. Subtle changes in the near and far fields to enable me to make a judgment as to what method to use (infinity or hfd). Further I wanted to test the idea of HFD focus for very close distances to see how sharp the image actually looked. This exercise helps in the composition of the image and gives me faith that when I want something to be in focus it will be!! I tried various CoC's and F# (in the camera range). I know that most people would have done similar exercises but I really have to do it myself.

Now about the Pictures.

These are a small set from all the pics that I took.

The last picture is to look at the subtle differences at the edge of the frame and the degree of focus close in.

The pictures with the tape measure are designed to show the degree of definition at the hfd/2. Where one sees the target (red tip) it is at the HFD and the camera lens is hfd/2 from the tape measure. The camera was set to faithful reproduction and no sharpening was employed.

I now understand!! After this I made a reference card to suite my lens and camera and I have faith that when I use it the results will be good.

How about that?

Jeremy.

Hi, Asher,

Asher Kelman said:

Doug,

Are you imagining the images or you can see them! Except for the first tiny ones, the others do not show up for me! However, you seem to be talking to ghosts that I can't see!

Click to expand...

I see them, I do see them!

I see them all "full size" in my post above.

Try this non-embedded URL:

http://lh4.ggpht.com/_nmcM5hLJalE/TJAfBfMGuMI/AAAAAAAAAEc/zfTwcJ1WeeA/cmp_hfd_f13_f16_4Mj.jpg

and then again here embedded:



Perhaps you haven't paid your Internet bill. Put another quarter in the Geyser.

Best regards,

Doug

Hi, Jeremy,

Jeremy Waller said:

Here's the last image Doug. When all 3 look good I'll write an explanation.

Click to expand...

Good enough. Thanks for the fix.

I prefer' em this size for this purpose:


And Asher can't see 'em at all!

Best regards,

Doug

Hi, Jeremy,

One critical parameter (you will doubtless cover this in your explanation) is what was the original frame size.

And of course what COCDL you used as the premise for your calculation of the HFD.

And what was that HFD.

Best regards,

Doug

Now I see them, all the way to the other side of the planet! At last I see!

IV'e payed my internet bill (I think) lol.

Here is the final result of this work - A reference card for the kit bag.

With a quick table:

And that abouts finishes this work!!

Jeremy.

Hi, Jeremy,

Jeremy Waller said:

IV'e payed my internet bill (I think) lol.
Here is the final result of this work - A reference card for the kit bag.

With a quick table:

Click to expand...

And now the Easy Eye version for us old geezers:


Thanks for the great work.

Best regards,

Dug

Doug Kerr said:

Hi, Jeremy,

One critical parameter (you will doubtless cover this in your explanation) is what was the original frame size.

And of course what COCDL you used as the premise for your calculation of the HFD.

And what was that HFD.

Best regards,

Doug

Click to expand...

Of course.

The original size of the image was 4368 x 2912 pixels (Canon 5D Mk 1) and the CoC was 2x8.2 microns (16.6 microns). This gave larger hfd's so there is some slack when setting up the camera.
I'll check my notes to see what the HFD was.

But in the mean time here is a better documented pic:

Jeremy

Jeremy Waller said:

With a quick table:

Click to expand...

Hi Jeremy,

Well done. Of course the differences per pixel will be almost invisible when the diffraction plays too much of a role. The transition between in- and out-of-focus is very gradual at these smaller apertures.

What does the "step" refer to?

Cheers,
Bart

Bart_van_der_Wolf said:

What is the "step" about?

Click to expand...

Hi Bart,

I dare guess that it is about how many steps does Jeremy have to walk to cover the required distance. It seems that if he takes 11 steps, he will cover 8 meters: i.e. a step span of 72 cm.

Cheers,

Cem_Usakligil said:

Hi Bart,

I dare guess that it is about how many steps does Jeremy have to walk to cover the required distance. It seems that if he takes 11 steps, he will cover 8 meters: i.e. a step span of 72 cm.

Click to expand...

Hi Cem,

I think so as well, but maybe our guesses are wrong? Besides, to get it down to 0.1 step accuracy, one would need a piece of string tied between one's legs ;-) (just kidding).

Cheers,
Bart

Hi, Jeremy,

Jeremy Waller said:

Of course.

The original size of the image was 4368 x 2912 pixels (Canon 5D Mk 1) and the CoC was 2x8.2 microns (16.6 microns). This gave larger hfd's so there is some slack when setting up the camera.
I'll check my notes to see what the HFD was.

But in the mean time here is a better documented pic:

Click to expand...

Thanks.

And for those who like the pix a little bigger:

Best regards,

Doug

Guys,

Why f 13-14? At this tiny aperture there's image degradation from diffraction. Do you really need to go less then f 8. With the 5D and 5DII I try to use f5.6.

Asher

Asher Kelman said:

Guys,

Why f 13-14? At this tiny aperture there's image degradation from diffraction. Do you really need to go less then f 8. With the 5D and 5DII I try to use f5.6.

Click to expand...

Hi Asher,

I can only guess that it's the result of trying to get the front of the HyperFocal DOF zone acceptably close.

There are 2 important criteria to balance.
1. The required DOF for the scene at hand, a more artistic consideration.
2. The required output enlargement while achieving a certain level of detail/resolution, a more technical consideration.

One can only achieve adequate (which can be anything from razor thin to infinite) DOF in a single shot by manipulating the aperture (or by using tilt). The technical requirement may require high quality (deconvolution) sharpening or focus stacking techniques is DOF is not enough when one tries to avoid deterioration of micro detail by diffraction. The 5d in this case will lose per pixel detail at apertures that are narrower than f/9 (based on my very strict criterion of 1.5x sensel pitch diffraction pattern diameter).

Cheers,
Bart

Hi, Bart,

Bart_van_der_Wolf said:

. . .(based on my very strict criterion of 1.5x sensel pitch diffraction pattern diameter).

Click to expand...

For what it's worth, that is in the neighborhood of the reported resolution (lines, not line pairs) for either the EOS 5D or 5D Mark II.

This corresponds to a Kell factor of about 0.67, a typical value predicted by Kell (1934)!

Best regards,

Doug

Hello Everybody,

Apologies for the late reply - back after a "mega health check".

I'll try to respond to everything.

1. Yes Bart the unit of "Step" is 0.75M one of my unhurried steps. I settled on this after "she who must be obeyed" refused to step out with my tape measure in public places. Using laser range finders (Bosch make very nice miniature ones) seemed too "geeky" so "Stepping it out" seemed most discreet - and it works quite well , lol.

Re:
"Well done. Of course the differences per pixel will be almost invisible when the diffraction plays too much of a role."

As we say here "for sure". Interestingly in this area I would like forums' opinion on the performance of what appears to a rather nice lens under some circumstances but absolutely terrible under other circumstances - The canon 85mm F1.2 lens. I only use it for low light level pictures. I collect my experimental data and post it in the lens section.

2. Doug, the F13/F14 (1/3 f=stop) was me not being able to make up my mind. I normally use F14.
The 5D only allows those 1/3 f-stop increments. Now being the engineering type I mainly use the F-stops in my table because they work so well. I'll try the wider (F10 and lower) a little later as the weather is actually improving and I'll be able to take pictures in comfort.

3. One of the ideas with these experiments was to learn to compose pictures with wide angle lenses and learning to make an informed compromises about sharpness in the near field and far field - I think I've got the hang of it!!

I'll now read the rest of the responses and see if I need to say more!!!

Kind Regards,

Jeremy.

Bart_van_der_Wolf said:

Hi Asher,

I can only guess that it's the result of trying to get the front of the HyperFocal DOF zone acceptably close.

There are 2 important criteria to balance.
1. The required DOF for the scene at hand, a more artistic consideration.
2. The required output enlargement while achieving a certain level of detail/resolution, a more technical consideration.

One can only achieve adequate (which can be anything from razor thin to infinite) DOF in a single shot by manipulating the aperture (or by using tilt). The technical requirement may require high quality (deconvolution) sharpening or focus stacking techniques is DOF is not enough when one tries to avoid deterioration of micro detail by diffraction. The 5d in this case will lose per pixel detail at apertures that are narrower than f/9 (based on my very strict criterion of 1.5x sensel pitch diffraction pattern diameter).

Cheers,

Bart

Click to expand...

My 2 Bob's worth:

For the 5D Mk1 the F# corresponding to the pixel spacing is in the vicinity of F11 - F12. I don't know if this is a good call as I naively assume all colours fall within 2 pixels perhaps 4 pixel CoC diameter would be a better call for the normal Bayer patterns (?).

Yes I agree with point 1. Artistic consideration (composition and degree of sharpness) is important for me.

Can't answer point #2 !!

Re: (or by using tilt)

I haven't reached that degree of sophistication. LoL.

Regards,

Jeremy

Jeremy Waller said:

For the 5D Mk1 the F# corresponding to the pixel spacing is in the vicinity of F11 - F12. I don't know if this is a good call as I naively assume all colours fall within 2 pixels perhaps 4 pixel CoC diameter would be a better call for the normal Bayer patterns (?).

Click to expand...

Hi Jeremy,

IMHO it is hard to determine a universal value for the COC that will suit everybody under all possible shooting/viewing conditions. Yes, there is a Bayer CFA, and there usually is an Anti-Aliasing filter, that reduces the contrast of microdetail, but there is also the influence of the Raw conversion which tries to undo some of the losses.

Nevertheless, on the majority of images that I specifically made to determine optimal response, I keep hitting the same threshold after which the per pixel micro-contrast deteriorates. It always is when the diffraction pattern diameter (for green light, say 555 nm) exceeds 1.5x the sensel pitch. The only exception is with residual lens aberrations, when (mostly) corner performance of a lens improves despite increasing diffraction, especially visible in the center.

That doesn't mean that a diffraction pattern diameter of 2x the sensel pitch suddely makes images useless, on the contrary. With good (deconvolution) sharpening a lot of the deterioration can be restored. But the pivot point is near 1.5x the sensel pitch, so I personally use that, knowing that it represents the onset of diffraction induced deterioration (with some maneuvering room to recover when stopped down further). So when I want to get the best possible resolution, I'll consider if a focus stacking scenario is warranted, or a somewhat diffraction limited startingpoint is acceptable (because I have tools for good sharpening, or a lens benefits from reducing residual aberrations, or I don't need to magnify the image).

Those are just the technical considerations, they follow the artistic decisions that preceded them.

Cheers,
Bart

Bart_van_der_Wolf said:

Hi Jeremy,

IMHO it is hard to determine a universal value for the COC that will suit everybody under all possible shooting/viewing conditions. Yes, there is a Bayer CFA, and there usually is an Anti-Aliasing filter, that reduces the contrast of microdetail, but there is also the influence of the Raw conversion which tries to undo some of the losses.

Nevertheless, on the majority of images that I specifically made to determine optimal response, I keep hitting the same threshold after which the per pixel micro-contrast deteriorates. It always is when the diffraction pattern diameter (for green light, say 555 nm) exceeds 1.5x the sensel pitch. The only exception is when (mostly) corner performance of a lens improves despite increasing diffraction, especially visible in the center.

That doesn't mean that a diffraction pattern diameter of 2x the sensel pitch suddely makes images useless, on the contrary. With good (deconvolution) sharpening a lot of the deterioration can be restored. But the pivot point is near 1.5x the sensel pitch, so I personally use that, knowing that it represents the onset of deterioration (with some maneuvering room to recover when stopped down further). So when I want to get the best possible resolution, I'll consider if a focus stacking scenario is warranted, or a somewhat diffraction limited startingpoint is acceptable (because I have tools for good sharpening, or I don't need to magnify the image).

Those are just the technical considerations, they follow the artistic decisions that preceded them.

Cheers,
Bart

Click to expand...

Hi Bart,

Yes determining the coc is hard because of value judgments - What is acceptable sharpness.

This is my final picture in this experiment where I try to judge what is a "good" picture of the room. When composing the picture questions came to mind: How important is the potted fern in the foreground? If it was not sharp then what? Even though it is close it should not become the subject of the picture.


About the de-convolution problem - that's really neat stuff! Those image processors do things that seem real "snake oil" but under certain conditions it works!! The thing that got to me was these chaps can reconstruct the phase of an image by knowing the magnitude only, and it works whats more. Nice stuff to work with - but currently I'm on the hook to complete a video quality exercise and comparison.

Regards,

Jeremy.

Hi, Jeremy,

Jeremy Waller said:

This is my final picture in this experiment where I try to judge what is a "good" picture of the room. When composing the picture questions came to mind: How important is the potted fern in the foreground? If it was not sharp then what? Even though it is close it should not become the subject of the picture.

Click to expand...

Thanks again for these useful test shots.

For the benefit of those who actually want to look at them:


You really should figure out how to reference the "full size" images!

Best regards,

Doug

The middle picture seems best as the fern plant has an attractive bowl and enough of both are well seen. However the scrappy edges of the plant on the left are distracting.

"Sharp" and "uniformly-sharp" are perhaps the most overused esthetics. However there's nothing "bad" with that in themselves. It should, I'd argue, serve some purpose and not be merely what the lens can do.

My feeling is that allocating color, light, shade, contrast, movement and so forth is what gives the picture your fingerprints! Still, some folk like Bresson with just one lens, (that's it, LOL!) aimed the camera from the right angle and just the right time and then did no "photoshopping"....well a little perhaps!

Asher

Doug Kerr said:

Hi, Jeremy,


You really should figure out how to reference the "full size" images!

Best regards,

Doug

Click to expand...

Apologies Doug. Yes I'll learn to do that.

Yes Asher the plant on the left is scrappy. That's the sought of thing that happens if one looks through the view finder but does not see.

Re: " My feeling is that allocating color, light, shade, contrast, movement and so forth is what gives the picture your fingerprints! "

So true! this is where the Artist-Photographer comes to the fore.

Re: " Still, some folk like Bresson with just one lens, (that's it, LOL!) aimed the camera from the right angle and just the right time "

Sounds sooo easy. BUT ....

Regards,

Jeremy.