Getting test target at right angles to lens axis.

[NigelC]

Just realised the other day that my canon 100IS macro lens was almost out of warranty and I'd never really tested it (haven't actually used it that much - it was insurance replacement).

After a couple of attempts to set up test target and camera, viewing at 100% on screen it looks to me as if sharpness reduces going diagonally from centre to top right hand corner (at wide apertures). In other words if you look at top right corner and scroll either down RH edge, or R to L along top edge it becomes progressively sharper. Only becomes equally sharp aross the frame at f8 to f11. Test was done using 2 sec self timer in LV, distance from sensor plane to target about 0.8m. Because of the obvious possibility of target a) not being flat and b) not being at right angles to lens axis, ended up putting target on floor and using tripod as a copy stand, relying on a spirit level in the hot shoe to level camera. I still think, before I say there is a fault with the lens (or the alignment of the lens mount with the sensor plane for that matter), I need to be sure that the angles are right - any suggestions for how to do this (without a test rig!)?

[NigelC]

Should have added camera was focussed manually, IS switched off, used a lupe on the LCD X10.

[Ellis Vener]

Two methods,  both based o nthe same idea.
1) Use a Zig-Align (approx. $140.00) and live view  for ensuring absolute squareness of sensor to target. This will be the most precise method. Link: http://www.zig-align.com/zig-align/zig-align_Home_page.html

2) Get a mirror you know to be very flat and set it up the same way you'll set up your target. Once again, use live view. Basically want you want to see is the center of the lens directly in the center of the imaging area.

[NigelC]

Tried a different test set up with neighbours brick wall (although a plumb line suggest this may be leaning fractionally towards my property !). Looks like previous tests were showing diffrerences in film and subject planes so I conclude my 100mm is not obviously bad and therefore not worth sending back to Canon. Also not obviously better than (my) 70-200f4L IS set to 100mm, both at f4. Maybe I've got a particularly good zoom.

[Bart_van_der_Wolf]

Two methods,  both based o nthe same idea.
1) Use a Zig-Align (approx. $140.00) and live view  for ensuring absolute squareness of sensor to target. This will be the most precise method. Link: http://www.zig-align.com/zig-align/zig-align_Home_page.html

2) Get a mirror you know to be very flat and set it up the same way you'll set up your target. Once again, use live view. Basically want you want to see is the center of the lens directly in the center of the imaging area.

Hi Ellis,

Indeed, both will do the job. However, for non-repro lenses (which are assumed to be specifically well corrected to achieve a flat focus plane) one has to consider that the plane of focus is not necessarily a flat plane due to curvature of field in most lenses.

That's why I always recommend to test corner resolution by focusing in the particular corner under investigation. That also reduces the sensitivity to non-squared shooting scenario's and narrow depth of field at wider apertures. A mild curvature of field is usually not an issue when shooting three dimensional subjects (there is alway an intersection with the focal plane somewhere).

Cheers,
Bart

[Ellis Vener]

Hi Ellis,

Indeed, both will do the job. However, for non-repro lenses (which are assumed to be specifically well corrected to achieve a flat focus plane) one has to consider that the plane of focus is not necessarily a flat plane due to curvature of field in most lenses.

That's why I always recommend to test corner resolution by focusing in the particular corner under investigation. That also reduces the sensitivity to non-squared shooting scenario's and narrow depth of field at wider apertures. A mild curvature of field is usually not an issue when shooting three dimensional subjects (there is alway an intersection with the focal plane somewhere).

Cheers,
Bart

His question, the one I was answering, wasn't above curvature of the projected image plane, whether a lens axis is de-centered, or other mechanical & optical aberrations  - those are different individual lens specific issues. I understood Nigel's  query to  be how to make sure the sensor/film plane is parallel with a subject plane. To that end I suggestion one low tech and one higher tech method. Testing for the issues you raise are other beach tar balls.

[Bart_van_der_Wolf]

I understood Nigel's  query to  be how to make sure the sensor/film plane is parallel with a subject plane. To that end I suggestion one low tech and one higher tech method.

You did, and I agreed with the suggestions.

Having that established however may still lead to the wrong conclusions about whether a lens is good, or whether it should be exchanged within the grace period.

Testing for the issues you raise are other beach tar balls.

Indeed, although having them pointed out may avoid jumping to wrong conclusions for the OP's intended goal. Just focusing in the individual corners will reveal if the lens can produce well focused results to begin with, and how much worse the corners are compared compared to the centre, and how they are compared to eachother. That is simple to do, curvature of field is not, I agree.

Cheers,
Bart

[stever]

the 70-100 is very sharp, esp at 100mm and i wouldn't expect you to see a difference from the 100 on a brick wall test.  using Imatest i measure a resolution difference of about 10%, but the resolution of both is so high, i'm not sure what printed subject would ever show up the difference.

[NigelC]

Hi Ellis,

Indeed, both will do the job. However, for non-repro lenses (which are assumed to be specifically well corrected to achieve a flat focus plane) one has to consider that the plane of focus is not necessarily a flat plane due to curvature of field in most lenses.

That's why I always recommend to test corner resolution by focusing in the particular corner under investigation. That also reduces the sensitivity to non-squared shooting scenario's and narrow depth of field at wider apertures. A mild curvature of field is usually not an issue when shooting three dimensional subjects (there is alway an intersection with the focal plane somewhere).

Cheers,
Bart

Well Ellis is right, I was basically just looking for tips on getting the 2 planes aligned. In the end I just marked the extreme L & R edges of the field of view on the wall and measured the distance from each edge to the edge of the lens hood. Spirit level sort of deals with vertical alignment.

I don't think the Canon 100 particularly suffers from curvature of field but I will try what you suggest when I get round to testing the Zeiss 28/2 which I think maybe does. However it is a bit difficult to focus into the corner when using LV , magnifying the central section, without moving the tripod head to focus and recompose, which can in itself introduce inaccuracies at close distances. I suppose I'd have to keep LV at full frame and use the loupe in the corner.

[Bart_van_der_Wolf]

Well Ellis is right, I was basically just looking for tips on getting the 2 planes aligned. In the end I just marked the extreme L & R edges of the field of view on the wall and measured the distance from each edge to the edge of the lens hood. Spirit level sort of deals with vertical alignment.

For the purpose of alignment, I keep a mirror tile (they come in sets of 10, from a local DIY home decoration store) in my bag. It's very easy to get alignment that way because you can see exactly what you're doing.

I don't think the Canon 100 particularly suffers from curvature of field but I will try what you suggest when I get round to testing the Zeiss 28/2 which I think maybe does.

The Canon 100mm maco lens(es) is (are) quite good. However, at high magnification factors your DOF is verry narrow, which leaves little room for error, especially when used with wide apertures. I recently helped a fellow photographer with deciding whether his new lens was good, and it had 1 corner that was significantly less detailed compared to the other 3. It remained so after repeating the experiment (to rule out a glitch in technique like in focusing or camera shake), so the lens was returned and exchanged for another copy, which performed very well in all corners and in the center.

A 28/2 has much more DOF at common shooting distances than a 100/2.8, but it still helps to focus individually on each corner.
  

However it is a bit difficult to focus into the corner when using LV , magnifying the central section, without moving the tripod head to focus and recompose, which can in itself introduce inaccuracies at close distances. I suppose I'd have to keep LV at full frame and use the loupe in the corner.

Using 10x zoom in LV, and shifting that LCD preview area to a corner for manual focusing (or contrast detect focusing) is the way to do it.

Cheers,
Bart

[NigelC]

Using 10x zoom in LV, and shifting that LCD preview area to a corner for manual focusing (or contrast detect focusing) is the way to do it.

Cheers,
Bart

I didn't realise you could do that!

[Ellis Vener]

I didn't realise you could do that!

Once I discovered that technique for still life and architectural work I was so happy I went out and bought a 3X Hoodman loupe and camera mount.

[NigelC]

Still not really happy I have determined whether this lens is within limits. Previous test of a brick wall at about 3 metres not really much of  a test for a macro lens so reverted to newsprint target at c. 0.6 metre. The only problem is trying to et everything perfectly aligned with the test target flat on the floor. I haven't got a large enough mirror so using plumb line to check lens axis is vertical and spirit levels and coins to level target.
Just how much depth of field is there at open aperture at that distance? I don't have a DOF table.
What i'm getting at f2.8 is one corner tends to go off a bit but this could well be unevenness in the target.

[Bart_van_der_Wolf]

Just how much depth of field is there at open aperture at that distance? I don't have a DOF table.

Depending on your criterion for the COC, 100mm @f2.8 at 0.6 metres distance gives approx. 1.6 millimetres DOF (assuming a COC of 0.0096 mm and a sensel pitch of 6.4 micron) if you want uncompromised resolution at the pixel level. IOW, no room for alignment error. You'd better score a mirror somewhere, because plumb lining by eye is not accurate enough at these apertures and magnification factors, and it get's more critical as you get closer.

Cheers,
Bart

[NigelC]

Depending on your criterion for the COC, 100mm @f2.8 at 0.6 metres distance gives approx. 1.6 millimetres DOF (assuming a COC of 0.0096 mm and a sensel pitch of 6.4 micron) if you want uncompromised resolution at the pixel level. IOW, no room for alignment error. You'd better score a mirror somewhere, because plumb lining by eye is not accurate enough at these apertures and magnification factors, and it get's more critical as you get closer.

Cheers,
Bart

Yes you are spot on - I actually found a DOF table online which indicated somewhat less than 1mm for 5D2 withh 100mm at that distance/aperture - IOW with the standard of alignment I've been able to achieve my test has absolutely no validity in identifying lens errors. I did through trial and errror find a position where was able to get corners v.close to centre sharpness just looking at using LCD @ x10. I think its probably OK

[Ellis Vener]

While testing our tools is always a good idea it needs to be balanced with a wariness of the dreaded "paralysis from over-analysis" syndrome.  The zig-align system I mentioned earlier ( http://www.zig-align.com/zig-align/zig-align_Home_page.html ) really is a god send for when you need to have the sensor plane be absolutely square to a subject. it is very fast to work with, reliable and saves a lot of time. Zig-Align claims that at two feet with their system you can get within 90 arcseconds* of being parallel.  If you are physics oriented  there is lots to read at the zig-align  site, but the system is simplicity itself to use.

Ensuring the two planes are parallel requires some very fine moves at the end of the aligning process so  try using  a three way adjustable leveling base like the Manfrotto 3416 leveling base (http://www.manfrotto.com/levelling-base) on top of your tripod head or use the leveling base in lieu of a tripod head. The three way adjustment of the 338 / 2416 (If you have a Really Right Stuff PCL-1 panning clamp ( or similar) put that on top of the leveling base.

*Also see: http://en.wikipedia.org/wiki/Minute_of_arc