Focus shift tests with 75/1.4 lux @ 1.5m

[ampguy]

Advertisement (gone after registration)

same methodology as for 35/1.4 pre-asph (see blog for details), but distance to center object was 150 cm.

 

images from tests at 1.4 through f8 are here

[rob_x2004]

:confused:Puzzling setup.

[jaapv]

I agree. The Summilux is most Cheezit at 5.6.

[ampguy]

Hi Rob,

 

Would you have any feedback that would make future tests a bit less puzzling? The idea is to accurately focus on a given spot (The label on the Cheezit box), at open aperture (1.4) and then without moving focus, changing apertures downwards to 2, 4, 5.6, and 8 to notice focus and DOF behaviors, at the given distance which is constant.

 

:confused:Puzzling setup.

[ampguy]

Thanks Jappv. I am very pleased to have a highly Cheezit lens at f5.6

 

I agree. The Summilux is most Cheezit at 5.6.

[Guest mc_k]

You should state that you are using the distance markers as targets.

 

Looking at the original-size photos, I judged it to be slightly front-focusing at f/2 and slightly back-focusing at f/2.8, so it should be most accurate between those two stops. I judged the middle of the depth of field, rather than try to judge the best focus.

 

I don't think this is a very sensitive test, or that you have measured the amount of the shift. You might repeat it with smaller intervals. You might also shoot a ruler; always better to use at least two different kind of tests.

[ampguy]

Advertisement (gone after registration)

Thanks, I appreciate the feedback.

 

When one does these tests one has to pick an aperture to index from, or start the series from, and judge focus bracket results based on that aperture. I used 1.4 - wide open.

 

After focusing on the center marker at 150 cm from the cheez-it box, the focus ring was never touched. Only the aperture ring.

 

I agree with you, that this lens seems optimized for about f2 and is consistent with my field work.

 

The challenges with rulers are that to see the face of the ruler and have no steps, is you have to either do some trig to shoot the ruler at an angle (~45 the most common), and now your sensor and subject focus points are not parallel.

 

I do agree with smaller steps. If I do the test again, I'll either setup a custom stepless ruler setup, or go to ~2.00cm intervals.

 

You can also check my blog post, as I first did this type of testing on a 35 pre asph lux, before the 75/1.4 lux.

 

Also, as you look at the originals, you are doing 100-200% crops, if judging resolution, right?

 

You should state that you are using the distance markers as targets.

 

Looking at the original-size photos, I judged it to be slightly front-focusing at f/2 and slightly back-focusing at f/2.8, so it should be most accurate between those two stops. I judged the middle of the depth of field, rather than try to judge the best focus.

 

I don't think this is a very sensitive test, or that you have measured the amount of the shift. You might repeat it with smaller intervals. You might also shoot a ruler; always better to use at least two different kind of tests.

[Guest mc_k]

...

The challenges with rulers are that to see the face of the ruler and have no steps, is you have to either do some trig to shoot the ruler at an angle (~45 the most common), and now your sensor and subject focus points are not parallel.

...

Also, as you look at the originals, you are doing 100-200% crops, if judging resolution, right?

 

I found both of the tests worthwhile so thanks for posting.

 

Well, of course the ruler plane and sensor plane are not parallel, or the whole ruler would be equally in focus; I'm not sure what your reservations are there? The numbers on the ruler can be divided by the square root of 2 so the units will be correct.

 

Yes, you measure a focus shift between two apertures, but you did not apparently give any measurement.

 

I get the same result at any magnification: at f/2 "106cm" is slightly sharper than "114cm," thus a focus to the front; at f/2.8 "114cm" is slightly sharper than "106cm," thus a focus to the back. This is in the end a little subjective. It's not uncommon to set up a lens that shifts to be a little forward wide open.

[Guest mc_k]

"bracketing" here means you shoot the whole series several times as a check, refocusing after every series.

[ampguy]

not sure I understand, the reason for my using boxes with distances from sensor labeled is that the whole box front and label are parallel to the sensor.

 

If I laid down a ruler on the floor, and shot at an angle, the sensor and ruler are not parallel anymore and distance, even in a mild crop are different.

 

If I angle a ruler and shoot it straight on, the vertical markings might be readable, but then one has to calculate what the real distance is from the sensor (easy trig), but will be a different distance of focus reference.

 

In other words, for this test, and I might use different procedures in the future, I wanted to use a constant distance - 110 cm from sensor for 35 pre asph lux, and rebuilt sonnar J8, and 150 cm from sensor for 75/1.4 lux.

 

I wanted measure using some procedures and constants that Sean Reid for example, doesn't track. He justs ensures his indexed aperture is spot on with focus bracketing, I think he does wide open too, then simply close down and look at crops, where he sometimes sees decreased center resultion upon stopping down from say f2 to f8, then center resolution coming back, on a heavy shifting lens, while corners continually increase in resolution (indicating something else is going on, possibly field curvature). However, he doesn't measure his distances, other than roughly 4-6 feet or so. He does test multiple samples, 4 in one case.

 

I highly recommend you subscribe to his newsletter, it's a fee site, but well worth it. Thanks again for your feedback. I do have some triangle rulers and ideas to accomplish a stepless system, but right now, I'd be pleased with having used 2cm intervals.

 

 

I found both of the tests worthwhile so thanks for posting.

 

Well, of course the ruler plane and sensor plane are not parallel, or the whole ruler would be equally in focus; I'm not sure what your reservations are there? The numbers on the ruler can be divided by the square root of 2 so the units will be correct.

 

Yes, you measure a focus shift between two apertures, but you did not apparently give any measurement.

 

I get the same result at any magnification: at f/2 "106cm" is slightly sharper than "114cm," thus a focus to the front; at f/2.8 "114cm" is slightly sharper than "106cm," thus a focus to the back. This is in the end a little subjective. It's not uncommon to set up a lens that shifts to be a little forward wide open.

[ampguy]

Yes. But, you cannot tell in the VF which is most spot on until you view highly magnified results, so you bracket a dozen times, but each dozen times, you shoot the whole aperture range and then utilize the set that shows the highest resolution at the target (110cm was my target for the 35 lux and rebuilt J8, 150cm was my target for the 75/1.4 lux.

 

Several years ago I was trying to help out a guy on another forum who had 20 or 30 odd 50mm lenses. Virtually none of his shots were in focus, but sometimes you could see where the focus was, perhaps on his daughters ear, even though he was focusing on her eyes. Another person then magnified these out of focus images, and all kinds of silly conclusions were made on softness and sharpness based on hundreds of out of focus test shots. I had recommended focus bracketing to this tester, but he thought it was silly. I even told him that you could nudge a long throw lens in either direction, yet not see any change in the VF, but see resolution changes in the results, again, he thought I was nuts. Fortunately, Sean Reid picked up on this, that's why I subscribe to his reports.

 

 

"bracketing" here means you shoot the whole series several times as a check, refocusing after every series.

[Guest mc_k]

not sure I understand, the reason for my using boxes with distances from sensor labeled is that the whole box front and label are parallel to the sensor.

 

If I laid down a ruler on the floor, and shot at an angle, the sensor and ruler are not parallel anymore and distance, even in a mild crop are different.

 

If I angle a ruler and shoot it straight on, the vertical markings might be readable, but then one has to calculate what the real distance is from the sensor (easy trig), but will be a different distance of focus reference.

...

 

But most home focus tests are oblique shots of a ruler or a ruled chart. It doesn't matter if you are aiming at the center of the ruler 150 cm away, or the center of a box 150 cm away, you are trying to measure where the plane of best focus intersects the target. If you, for example, read off 5 cm back-focus from the ruler, then the back-focus is actually 5 / sq. rt. of 2. (This, from having shot the ruler at 45 deg.)

 

Again, if the ruler plane and sensor plane were parallel the entire ruler would be equally in focus. I think you are missing the basic point of a focus test, which is measuring the distance between two planes: the plane of intended focus, and the plane of best focus.

[Guest mc_k]

Also...confusion about two different kind of tests. In a resolution test, you should bracket by testing at slightly different distances and pick the best. In a focus test, you should bracket by testing multiple times at the same distance and average the error, i.e. not throw out any tests.

[ampguy]

Perhaps you could link me to the type of test you'd like to see and I'll explain why I prefer my methods. I do appreciate your input, and have seen some of these tests, but I'm trying to accomplish some things that they are not.

 

Are you referring to the DSLR bars that are placed on the carpet and shot at a ~45 angle downwards?

 

 

QUOTE=mckeough_k;1262906]But most home focus tests are oblique shots of a ruler or a ruled chart. It doesn't matter if you are aiming at the center of the ruler 150 cm away, or the center of a box 150 cm away, you are trying to measure where the plane of best focus intersects the target. If you, for example, read off 5 cm back-focus from the ruler, then the back-focus is actually 5 / sq. rt. of 2. (This, from the Pythagorean theorem, and having shot the ruler at 45 deg.)

 

Again, if the ruler plane and sensor plane were parallel the entire ruler would be equally in focus. I think you are missing the basic point of a focus test, which is measuring the distance between two planes: the plane of intended focus, and the plane of best focus.

[Guest mc_k]

Perhaps you could link me to the type of test you'd like to see and I'll explain why I prefer my methods. I do appreciate your input, and have seen some of these tests, but I'm trying to accomplish some things that they are not.

...

 

actually I have no preference for one kind of test over another. I suggested a "ruler" because in your 35mm test you said you wanted to use one next time.

 

You hinted above there is some kind of "distance" error involved in shooting a ruler at 45 degrees. (Because the ruler scale is now off? Because the viewfinder is offset from the taking lens? Something else?) ... Anyway, yes, I would be interested to hear what the advantage of your test is. Similar tests have been done here with staggered books, bottles, and suspended strings.

[Guest mc_k]

...you bracket a dozen times, but each dozen times, you shoot the whole aperture range and then utilize the set that shows the highest resolution at the target...

 

...I even told him that you could nudge a long throw lens in either direction, yet not see any change in the VF, but see resolution changes in the results, again, he thought I was nuts. Fortunately, Sean Reid picked up on this, that's why I subscribe to his reports.

 

The kind of error you describe is exactly why you shouldn't take the "highest resolution" result and throw the others out, if you are testing focus. You take the average of a large enough sample to minimize the effect of this kind of error.

[ampguy]

This set of tests is for one thing only, to determine focus shift of a lens, not to determine if the lens has front or back focus or anything else.

 

The critical focus at the indexed aperture (wide open for me) is very important, and focus bracketing examining results, is the only way to tell when the lens is optimally focused at the indexed aperture.

 

Depends how you shoot a ruler, but the most common ways I see it used are at angles where a specific mark is used for a target say the 5" mark on a foot ruler from 0" to 12", or the popular DLSR focus test where a set of lines with a middle "Focus Here" thick line is and marks above and below.

 

Those tests above are fine for casually determining front or back focus, but not for focus shift or for examining resolution.

 

The instructions for the DSLR test is to shoot the lined paper with your camera at a 45 degree angle, so if you examine detailed results, at the point of best focus, you're not viewing when your sensor (or film) were parallel to that target.

 

But back to focus shift, to get back to rulers and markers, markers are necessary, there is no way to do a stepless test, but many ways to use finer granularity than I used. Assume there is no DOF with focus shift testing, assume it is binary, you only want to see if the single point of best focus, that should be pre-determined with focus shifting, and that set of images stopped down, not having had the focus ring touched from the index (wide open) image.

 

With the above strict binary definition, all lens would have focus shift, but with my method you can see how it affects your images in the field. For example, with my 75/1.4 lux, I know that close up (< 150cm), I can safely focus on eyes, and sometimes ears, and get the front of the face and nose in focus at 1.4, while at 2 and 2.8 (again at 150cm) I can focus on precisely what I want in sharpest focus, or I can intentionally focus a couple of cm in front, to shift the whole DOF window (OK I did bring up DOF) up front a bit which will have an effect of shallower DOF (or moving the DOF window closer, if there were objects other than air between subject and sensor/film that I wanted in sharper focus).

[Guest mc_k]

but you still have not given any rationale why the sensor has to be "parallel to the target." As far as I know the current setup in Solms uses an oblique target, and the same for the past setup in NJ.

 

This was the topic of a long thread last year or the year before.

[ampguy]

It's just a personal preference or convenience for me.

 

If I am interested in seeing focus shift differences between 2 lenses, and am examining the shot at say f5.6 from > 100cm, and I want to zoom into a small object at several hundred magnification, I want to be assured that the top, bottom, left, and right were parallel to the sensor, and that the top wasn't closer than the bottom, or left closer than the right, or vice versa.

 

A lot of folks have sent their lenses to Leica and have still not been pleased with the results upon return. That's all I really want to say about that.

 

but you still have not given any rationale why the sensor has to be "parallel to the target." As far as I know the current setup in Solms uses an oblique target, and the same for the past setup in NJ.

 

This was the topic of a long thread last year or the year before.

[Advertisement]

Hallo,

Look here: Leica M

[Guest mc_k]

This set of tests is for one thing only, to determine focus shift of a lens, not to determine if the lens has front or back focus or anything else.

...

 

well since the test determines focus shift, how many cm focus shift did you get between f/1.4 and f/8 ? Again, I thought it was a wortwhile test, and seems to be lacking only your conclusions.