How to determine true focal length

[mdarnton]

To find the true focal length, focus at infinity and measure the distance from any part of the lens to the film. Then focus accurately at 1:1 and measure the same way. The focal length is the difference between the two distances. This may be harder than it sounds, but the theory is good.

 

[Nodda Duma]

I had understood that the entrance pupil could be inside the lens, to the front of the lens or behind the lens depending on lens design. I can't see how the eyepiece method will always work for inside or outside, especially if the entrance pupil is behind the eyepiece or behind the lens. Am I wrong ?

Just move the eyepiece to focus on the E.P. and note the distance, forwards or backwards.

 

[bernard_L]

@ mdarton. your proposed method assumes 1:1 magnification; but you don't give details on how to achieve that?
@ Nodda Duma. The OP wants to find the focal length, not the position of the focal plane
@ couldabin. Indeed the distance between the rear nodal point and the image focal plane is the focal distance (in the common case where input and output are both in air, and nodal points are in principal planes. But, a point is not a distance; in other words, if your follow that approach, you'll need to determine both the nodal points and the focal plane, whan what you really need is only the distance between these two.

Note in passing that this reconciles the criteria for the two kinds of pano cameras: (A) lens and slot rotate, film stationary in cylindrical slot, lens rotates around its rear nodal point; (B) design envisioned by OP, where whole camera rotates, and film motion must match image motion in focal plane; linear velocity of film = angular velocity of camera times focal length.

And... we see the old confusion between entrance pupil and front nodal point rearing its head again.

May I suggest a method that bypasses all intermediate steps (determine focal length, derive film velocity, implement that...), and provides a verification of proper operation of the as-built camera as a system. I assume that your camera is built, and the focusing issue has been solved, i.e. the film is in the focal plane. Replace your full-height vertical slot (behind which the film should be moving at just the right speed to avoid blurring), by two slots, each one spanning half of the image height, and separated laterally by a factor of a few times their individual width. Take a pano pic of a scene with vertical features (poles, trees, fence...), using the nominal focal length as a basis for film velocity. If the vertical features are aligned between top and bottom, you're done. If not, take a guess (or think hard) as to whether the film velocity should be increased or decreased. Once you have two pics with the misalignment in opposite directions, make a plot of misalignment versus film velocity (the plot has just two points and a straight line through them); interpolate to find the velocity for zero misalignment. Done.
All this assumes you can change the film velocity continuously and reproducibly. And if you can't, you are in a spot regardless of the method.

As a distinct issue, if you want to avoid parallax blurring between near and far objects, your camera should rotate around the entrance pupil of the lens. Practically, make that the "mechanical" front flange and you'll be pretty good.

 

[RobC]

Just move the eyepiece to focus on the E.P. and note the distance, forwards or backwards.

Thats easier said than done since you can't see it and probably don't know where it is anyway unless you have accurate way to determine that too.

 

[RalphLambrecht]

The flange focal length is relatively easy to determine empirically...

1. mount the lens on your 4x5 camera
2. focus on something far, far away
3. measure the distance between the flange (the front of the lensboard) and the film plane

That's it!

If the lens is modern and made by one of the respected manufacturers, then specfifications are usually published and available...one simply needs to ask or dig with google.

+1 and you cannot trust advertised scanner spi(samples per inch)

 

[Nodda Duma]

Thats easier said than done since you can't see it and probably don't know where it is anyway unless you have accurate way to determine that too.

Hmm...maybe you are confusing the entrance pupil with the physical aperture (ie the shutter blades )? They are two distinct though related entities in a lens. The entrance pupil is the *image* of the aperture as seen through the optics in front of it. The exit pupil is analogous but as seen through the optics behind the aperture. When I say focus on the aperture I could also say bring the eyepiece to focus on the shutter blades. Naturally you don't know where the entrance pupil is before you perform the measurement (otherwise why do so). Maybe that help clears it up?

An Internet forum is *the worst* platform for discussing technical details. If we were face-to-face I could easily demonstrate the method and in two minutes you would say "ah hah makes perfect sense!"

@ Bernard_L: the (effective) focal length *is* the distance to the focal plane under certain test conditions...conditions described in the link in my post. They are examples of how the lens effective focal length is measured/verified professionally.

 

[bernard_L]

@ Nodda Duma

@ Bernard_L: the (effective) focal length *is* the distance to the focal plane under certain test conditions...conditions described in the link in my post. They are examples of how the lens effective focal length is measured/verified professionally.

The distance to the focal plane... from what? Need two points to define a distance. Which is precisely what I state my previous post: "the focal distance (in the common case where input and output are both in air, and nodal points are in principal planes)"
Plus: "The OP wants to find the focal length, not the position of the focal plane"
Plus: The business about the entrance pupil is a separate issue, to avoid parallax errors between near and far objects. That defines what should be the rotation axis of the camera body (design type B in my previous post). But that is not the question of posed by couldabin. He needs to know the focal length (irrespective of the position of the nodal point or the focal point) to determine the rate of film displacement as the camera rotates. Design type (B) allows a 360° panoramic; design type (A) maxes out around 180°.

Before anyone mentions nodal points again, please read:
https://en.wikipedia.org/wiki/Panoramic_photography#Panoramic_cameras_and_methods
that clearly explains the difference between "short rotation" (what I call type A) and "full rotation" (what I call type B).
The OP has clearly identified his need to convert an angular (apparent) displacement of the scene into a linear displacement in the focal plane, to be matched by the film motion. The focal length, and only the focal length, is needed to relate these two quantities.

 

[Nodda Duma]

Bernard,

I'm afraid you didn't actually read my post before you nitpicked what I posted.

@ Nodda Duma


The distance to the focal plane... from what? Need two points to define a distance.

And in my original post I was very precise in what I typed, which was flange focal length. This is a very specific definition which does measure from a specific point in the optical system. I also mentioned effective focal length, which is also a very specific term. The definitions are easy to find, as well as information on why nodal points are important for the original inquiry. I agree...the information he needs can be derived from any of the several different measurements of focal length. Honestly, I'm not even sure of the point you're trying to make other than "don't post information that I am not familiar with."

Of course you are also correct...this thread diverged a fair amount from the OP and touches on topics including finding the entrance pupil. The EPP is actually related to his request. I also don't think there is anything wrong in replying with helpful info (that the OP or anyone else reading may be interested in).

Truly, with all due respect,
Jason

 

[couldabin]

May I suggest a method that bypasses all intermediate steps (determine focal length, derive film velocity, implement that...), and provides a verification of proper operation of the as-built camera as a system. I assume that your camera is built, and the focusing issue has been solved, i.e. the film is in the focal plane. Replace your full-height vertical slot (behind which the film should be moving at just the right speed to avoid blurring), by two slots, each one spanning half of the image height, and separated laterally by a factor of a few times their individual width. Take a pano pic of a scene with vertical features (poles, trees, fence...), using the nominal focal length as a basis for film velocity. If the vertical features are aligned between top and bottom, you're done. If not, take a guess (or think hard) as to whether the film velocity should be increased or decreased. Once you have two pics with the misalignment in opposite directions, make a plot of misalignment versus film velocity (the plot has just two points and a straight line through them); interpolate to find the velocity for zero misalignment. Done.
All this assumes you can change the film velocity continuously and reproducibly. And if you can't, you are in a spot regardless of the method.

OK, I'm really interested in this test approach, because so far home-constructed testing has shown the lens to anything but the specified focal length. But I'm having trouble visualizing it (and therefore why it works). Two half-height slits, separated by a distance of, say, 0.200", assuming they each are 0.06" wide. Does one extend from the top to the middle, and the other from the middle to the bottom?

Correct me if I'm wrong, but I believe that pivoting around anything but the rear nodal point will cause the image to creep, something that of course can be compensated for but also a phenomenon that allows for correcting film speed error. So eliminating parallax error will necessarily involve compensating for this, right?

 

[bernard_L]

Does one extend from the top to the middle, and the other from the middle to the bottom?

Correct.
Optionally (here I assume 120-size film) one slot 35mm high from the top, one 35mm from the bottom, so their imaging overlaps over 2x35-56=14mm; in that region, if the camera is misaligned, a vertical object will have a double image. Kind of like a hybrid between a split image center spot in a SLR and a superposition double-image spot in a rangefinder.

Correct me if I'm wrong, but I believe that pivoting around anything but the rear nodal point will cause the image to creep

Let's double-check that I'm on the same page as you. You stated as a response to my first post: "The panoramic camera will have a film transport that moves the film as the camera body rotates". Please confirm that indeed you are indeed designing a pano-camera of the full-rotation type in the sense of https://en.wikipedia.org/wiki/Panoramic_photography#Panoramic_cameras_and_methods (e.g. something like a Cirkut) and not of the short-rotation type (e.g. Noblex, Horizon). So, your camera rotates as whole, while, inside the camera, film advances behind a slit at a TBD speed.

Then, let's tackle the problems one at a time.

• Scene is at infinity (landscape). Your problem is to have the film move at the slit location at precisely the same speed as the image does (relative to the camera body), due to the rotation of the camera relative to the scene. I understand that is the subject of your original post. The answer being v=fxA, where v is the linear velocity (say, mm/s) of the film, f the focal length (say, in mm), and A the angular velocity of the camera rotation (radian/sec). Of course you know that since you ask about the focal length.
• Part of scene is at finite distance. Need to avoid parallax shift of near versus far objects as the camera swings. (a) need to recognize this is a distinct issue from the previous one, or the discussion becomes confused; (b) technically correct way to address this is to have the camera body swing around the entrance pupil of the lens; (c) that issue is quantitatively much less important than when assembling a panoramic image from discrete pictures: in your case, any part of the scene is captured only over a fairly narrow range of camera angles (in relation to the width of the slit), while in an assembled panoramic, one needs consistency between (say) the left side of one image and the right side of the next one, taken typically 30° apart. Assuming the slit width you quote "assuming they each are 0.06" wide" is typical, that is 1.5mm, and a 90mm f.l., any part of the scene is seen by the film only over a range of camera rotation of 1.5/90=0.0167 radian ~0.95°. So, as concerns issue (c), and pending a more quantitative analysis I'd say that it's better to have the axis of rotation closer to the front than to the back of the camera, but no need to agonize over that issue.

May I suggest that during the initial design, you ask yourself not only what is the value of this or that design parameter (focal length, film velocity), but also, what is the tolerable error that will still allow a successful camera against some criterion (like, circle of confusion, blurring length, etc). So you know where you must put your effort.

A question out of curiosity: do you intend to achieve film motion by mechanical coupling to the camera rotation (with suitable gear ratio) or electronically, or???

 

[bernard_L]

Just realized I had not answered that:

But I'm having trouble visualizing it (and therefore why it works).

Well, the film velocity equation v=fxA (my previous post) ensures that, while the film is exposed (traveling across the slit width) it is stationary with respect to the image. Any other velocity will result in the image "creeping" relative to film during its travel over the slit width, and therefore image blur. But, errors in film velocity of equal magnitude and opposite signs result in just the same amount of blurring: not very helpful to converge on the correct value. The purpose of the two separate slits is to amplify this creep (separation between slits larger than width). If film velocity is not correct, image creeps with respect to film during the time it takes for film to travel from slit #1 to slit #2. So, a telegraph pole will appear "broken" where the two slits meet height-wise.

As I stated in my post of yesterday, you can either think hard and deduce from the "break" in the telegraph pole (upper half: does it appear displaced to the right or to the left of the bottom part) the direction sign and magnitude of the correction to the film velocity (slower, faster, by how much). Or you can deduce the correct velocity from "failed" experiments, ideally a pair of experiments with errors of opposite signs (again, see my initial description of the proposed method)

I also repeat that the advantage of using the complete camera for adjustment is that all possible errors are taken into account and tested: focal length, film advance drive, camera rotation drive, etc. You don't need to know each one to full accuracy, as long as the v=fxA equation is obeyed.

 

[couldabin]

Thanks for this. The two-slit test is just brilliant -- simple and effective. And I'm very partial to any solution that's easily implemented by DIYers. This will move the project forward immensely.

 

[couldabin]

Correct.
Optionally (here I assume 120-size film) one slot 35mm high from the top, one 35mm from the bottom, so their imaging overlaps over 2x35-56=14mm; in that region, if the camera is misaligned, a vertical object will have a double image. Kind of like a hybrid between a split image center spot in a SLR and a superposition double-image spot in a rangefinder.


Let's double-check that I'm on the same page as you. You stated as a response to my first post: "The panoramic camera will have a film transport that moves the film as the camera body rotates". Please confirm that indeed you are indeed designing a pano-camera of the full-rotation type in the sense of https://en.wikipedia.org/wiki/Panoramic_photography#Panoramic_cameras_and_methods (e.g. something like a Cirkut) and not of the short-rotation type (e.g. Noblex, Horizon). So, your camera rotates as whole, while, inside the camera, film advances behind a slit at a TBD speed.

Then, let's tackle the problems one at a time.

• Scene is at infinity (landscape). Your problem is to have the film move at the slit location at precisely the same speed as the image does (relative to the camera body), due to the rotation of the camera relative to the scene. I understand that is the subject of your original post. The answer being v=fxA, where v is the linear velocity (say, mm/s) of the film, f the focal length (say, in mm), and A the angular velocity of the camera rotation (radian/sec). Of course you know that since you ask about the focal length.
• Part of scene is at finite distance. Need to avoid parallax shift of near versus far objects as the camera swings. (a) need to recognize this is a distinct issue from the previous one, or the discussion becomes confused; (b) technically correct way to address this is to have the camera body swing around the entrance pupil of the lens; (c) that issue is quantitatively much less important than when assembling a panoramic image from discrete pictures: in your case, any part of the scene is captured only over a fairly narrow range of camera angles (in relation to the width of the slit), while in an assembled panoramic, one needs consistency between (say) the left side of one image and the right side of the next one, taken typically 30° apart. Assuming the slit width you quote "assuming they each are 0.06" wide" is typical, that is 1.5mm, and a 90mm f.l., any part of the scene is seen by the film only over a range of camera rotation of 1.5/90=0.0167 radian ~0.95°. So, as concerns issue (c), and pending a more quantitative analysis I'd say that it's better to have the axis of rotation closer to the front than to the back of the camera, but no need to agonize over that issue.

May I suggest that during the initial design, you ask yourself not only what is the value of this or that design parameter (focal length, film velocity), but also, what is the tolerable error that will still allow a successful camera against some criterion (like, circle of confusion, blurring length, etc). So you know where you must put your effort.

A question out of curiosity: do you intend to achieve film motion by mechanical coupling to the camera rotation (with suitable gear ratio) or electronically, or???

Yes, the camera I'm building will be as you described -- a body/lens unit that rotates, with 120 film (sans backing paper) being advanced at the appropriate rate. I toyed with having the film be stationary, mounted on a circular support but went with this other approach because a) I couldn't see how to get a full 360 shot; b) the logistics of starting/stopping the exposure seem daunting; and c) the film would have to be loaded/unloaded in a changing bag or darkroom. So coming up with a film transport, which itself poses a number of challenges, solves all three of the the above concerns and lets me use my 4x5 lens. (If this really does work, I may look for a medium format lens with better specs.) I'm starting with mechanical coupling -- I have a small lathe/mill that makes it pretty practical to machine custom pulleys, and I have in mind using a combination of gears and one final belt/pulley that will allow for fine-tuning transport speed. I have no idea how close actual film velocity will be to theoretical but am hoping that with the right choice of belt material (I see that replacements for record players/VCRs/etc are readily available in a rich combination of sizes/lengths/styles) it will produce repeatable rates. Being able to compensate by moving the pivot point makes it all the more attractive.

Of course, that compensation affects parallax, but I now see that a person could build that creep into the transport calculation, allowing for a pivot point that is nominally the entrance pupil (but deviated by the amount necessary to correct for mismatches in the final transport speed. Is the "creep" caused by pivoting about a point other than the rear node basically (offset distance * 2 * pi) per revolution?

The basic design philosophy is to come up with something using as many on-hand objects as possible, just to determine who viable this is. If it works well (which, I'll admit, is better than I'm expecting) I will be looking for a 12V variable speed motor to provide the rotation. Going with stepper motors for the rest puts this into an entirely different price range. Right now, I'm using time that isn't making me any money anyway ...