This action might not be possible to undo. Are you sure you want to continue?
∗
Robert E. Wheeler
May 8, 2003
∗
c 19972001 by Robert E. Wheeler, all rights reserved.
1
Contents
1 Desargues’s Theorem 4
2 The Gaussian Lens Equation 6
3 Thick lenses 8
4 Pivot Points 9
5 Determining the lens tilt 10
5.1 Using distances and angles . . . . . . . . . . . . . . . . . . . . . . 10
5.2 Using back focus . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
5.3 Wheeler’s rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
5.4 Lens Movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
5.5 Back Tilts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
6 Depth of ﬁeld for parallel planes 15
6.1 Near DOF limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
6.2 Far DOF limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
6.3 DOF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
6.4 Circles of confusion . . . . . . . . . . . . . . . . . . . . . . . . . . 18
6.5 DOF and format . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
6.6 The DOF equation . . . . . . . . . . . . . . . . . . . . . . . . . . 19
6.7 Hyperfocal distance . . . . . . . . . . . . . . . . . . . . . . . . . 20
6.8 Approximations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6.9 Focus given near and far limits . . . . . . . . . . . . . . . . . . . 21
6.9.1 Object distances . . . . . . . . . . . . . . . . . . . . . . . 21
6.9.2 Image distances . . . . . . . . . . . . . . . . . . . . . . . . 22
7 Depth of ﬁeld, depth of focus 23
8 Fuzzy images 24
9 Eﬀects of diﬀraction on DOF 26
9.1 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
9.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
9.3 Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
9.4 Format considerations . . . . . . . . . . . . . . . . . . . . . . . . 31
9.5 Minimum aperture . . . . . . . . . . . . . . . . . . . . . . . . . . 32
9.6 Theoretical curves . . . . . . . . . . . . . . . . . . . . . . . . . . 33
10 Depth of ﬁeld for a tilted lens 35
10.1 Near and far DOF equations . . . . . . . . . . . . . . . . . . . . 35
10.2 Near and far DOF equations in terms of ρ . . . . . . . . . . . . . 35
10.3 Focus given near and far limits . . . . . . . . . . . . . . . . . . . 37
10.3.1 Object distances . . . . . . . . . . . . . . . . . . . . . . . 37
2
10.3.2 Image distances . . . . . . . . . . . . . . . . . . . . . . . . 38
10.4 The graph of the DOF limits . . . . . . . . . . . . . . . . . . . . 38
10.5 Variation of maximum diameter of the ellipse of confusion with ρ 39
11 Image control 41
11.1 Exposure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
11.2 Light meters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
11.3 18% gray cards . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
11.4 Sunny 16 rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
11.5 Contrast control — the Zone system . . . . . . . . . . . . . . . . . 43
11.6 Characteristic curves . . . . . . . . . . . . . . . . . . . . . . . . . 44
11.6.1 Graduating function . . . . . . . . . . . . . . . . . . . . . 44
11.6.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
12 Perspective and normal lenses 46
13 Digital 48
13.1 Scanning and resolution . . . . . . . . . . . . . . . . . . . . . . . 48
13.2 Digital cameras . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
14 Miscellany 50
14.1 Fractional stops . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
14.2 Stopping Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
14.3 Bellows Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
14.4 Stops as distances . . . . . . . . . . . . . . . . . . . . . . . . . . 52
14.5 AOV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
14.6 Film ﬂatness and vibration eﬀects . . . . . . . . . . . . . . . . . 53
14.7 Hand holding a camera . . . . . . . . . . . . . . . . . . . . . . . . 53
14.8 Finding the 1:1 focus point . . . . . . . . . . . . . . . . . . . . . 53
14.9 Miscellaneous equations . . . . . . . . . . . . . . . . . . . . . . . 53
3
Preface
The following is a collection of notes and derivations that I made while reading
about view cameras. Everything is geometrical (trigonometrical), and thus may
not apply in various ways to actual lenses; although it should be pointed out
that while actual lens’s may not attain these ideals, it is these for which lens
designers strive.
1 Desargues’s Theorem
There are three fundamental planes: the ﬁlm plane, FP, the lens plane, LP,
and the subject plane, SP. An object in SP will be in focus on FP only when
the three planes bear a particular relationship. This relationship, interestingly
enough, may be derived from a theorem of Desargues (15931662) in projective
geometry.
Projective geometry is the name for that branch of mathematics that arose
to explain the relationships of three dimensional objects, projected onto a two
dimensional canvas; these relationships were observed by artists in the Renais
sance (about 1400) when they began to develop an understanding of perspec
tive. To draw a rug on a ﬂoor, for example, whose sides appear proper, one
must choose a vanishing point, VP, on the horizon at which lines through the
sides of the rug converge. The interesting thing is that if one draws a object, say
box in three quarter view, then lines through its sides must converge to VP’s
on the horizon and all such VP’s must lie on the same line.
This convergence to points on a line is the essence of Desargues’s theorem.
Theorem (Desargues): If in a plane two triangles ABC and
abc are situated so that the straight lines joining corresponding ver
tices are concurrent in a point O, then the corresponding sides, if
extended, will intersect in three collinear points.
This is illustrated in space by ﬁgure 1. The theorem is rather diﬃcult to prove
for triangles lying in the same plane, but it is quite easy to prove if they lie
in diﬀerent planes. In ﬁgure 1, aA, bB, cC are concurrent at O according to
the hypothesis. Now ab lies in the same plane as AB, so that these two lines
intersect at some point P2; likewise ac and AC intersect at P3, and cb and CB
intersect at P1. Since P1, P2, and P3 are on extensions of the sides of abc and
ABC, they lie in the same plane with each of these two triangles, and must
consequently lie on the line of intersection of these two planes. Hence P1, P2,
and P3 are collinear which was to be proved.
4
O
a
c
b
A
C
B
P1
P2
P3
Figure 1: Desargues’s conﬁguration in space
It should be noted that the theorem holds, and the above proof follows
through if one places the concurrent point O at inﬁnity, as illustrated by ﬁgure
2, which brings us to Photography, since the three lines SP, LP, and FP in ﬁgure
2 represent the three photographic planes mentioned above. P1 may also be at
inﬁnity, in which case the three planes are parallel.
The point b in ﬁgure 2 is the lens center, and the lines P2a, and P3c are
rays through the lens from SP to FP. The triangles P2Aa and P3Cc are those
usually used to graphically depict image formation — see Stroebel, et.al. (1986).
The lines aA and bB
1
and cC are drawn to be orthogonal to LP. Hence the
extensions of aA, cC and bB are coincident at a point at inﬁnity, satisfying
the condition of Desargues’s theorem, and thus the points P1, P2, and P3 are
collinear. Since such a diagram may be drawn for any three lines which meet
at a point, one has Scheimpﬂug’s rule.
Scheimpﬂug’s rule: The subject plane is rendered sharp when
the subject plane, the lens plane, and the focus plane all meet in a
line.
1
B may be placed anywhere on the line orthogonal to LP as far as Desargues’s theorem is
concerned, but for Photography, it should be placed one focal length away from b, making it
the object focal point.
5
P1
P2
P3
O
B
A
C
a
b
c
SP
LP
FP
Figure 2: The three planes
2 The Gaussian Lens Equation
Figure 3 shows the three planes from ﬁgure 2 with some additional lines and
points. The lines Aa, Bb and Cc are orthogonal to LP. The point L is the lens
center, and the dashed line parallel to SP ending at point c is a line from inﬁnity:
since this line meets SP at inﬁnity Desargues’s theorem is satisﬁed. The length
of Cc, f, is deﬁned, as is customary, as the focal length of the lens. As has been
noted, the three planes satisfy the Scheimpﬂug rule. We will now show that
Desargues’s theorem also implies the Gaussian lens equation:
1
u
+
1
v
=
1
f
. (1)
Using ﬁgure 3, equation (1) may be obtained as follows:
1
u
+
1
v
=
1
Pb sin α
+
1
Pa sin β
,
=
1
Pb
f
cL
+
1
Pa
f
Pc
,
=
1
f
cL
Pb
+
Pc
Pa
,
and since by similar triangles cL/Pb = ca/Pa, one has
1
u
+
1
v
=
1
f
ca
Pa
+
Pc
Pa
=
1
f
.
6
One can express this in terms of distances along the line ab as
1
h
+
1
y
=
cosω
f
(2)
where u = hcosω and v = ycosω.
a
SP LP
FP
v
u
A
b
B
f
C
c
α
β
α
L
P
h
y
ω
Figure 3: The lens equation
7
There are several derived forms of equation (1) that are frequently of use.
u =
vf
v −f
, (3)
v =
uf
u −f
. (4)
And if one deﬁnes magniﬁcation by
m =
v
u
=
v −f
f
=
f
u −f
, (5)
then
u = f(1 +
1
m
), (6)
v = f(1 +m), (7)
and of course the Newtonian form:
f
2
= (u −f)(v −f). (8)
These equations hold for h and y when f is replaced by f/cosω as per
equation(2). Later, use will be make of the fact that magniﬁcation, m, is con
stant for all values of ω. This follows algebraically and by inspection of the
similar triangles in ﬁgure (3).
3 Thick lenses
The above is geometrical, and is an approximation to a thin lens, which is an
abstraction describing a lens such that, within acceptable precision,the thickness
can be ignored. Camera lenses are not thin. Fortunately, the consequences are
negligible.
Light from a lens may be described geometrically as if it radiated from
a single surface, which is essentially a plane for paraxial rays called the rear
principal plane. This plane acts like a plane though a thin lens, and the focal
length, distance v, and other variables on the lens side act as in the geometric
theory above. The geometry is attributed to Gauss, see (Kingslake 1978) or
(Mouroulis and Macdonald 1997). Because the lens is thick, variables on the
object side are not quite what is given by the thin lens geometry, but the
diﬀerences are usually so small as to have no eﬀect on the acceptable precision
of the calculations. The front principal plane is cognate to the rear principal
plane in that they are images of one another with unit magniﬁcation. If δ is
the nodal space
2
, then u becomes u
= u − δ/ cos α ≈ u − δ, where α is the
2
The nodal space or hiatus is the thickness of the lens as measured by the diﬀerence between
the two nodal points. In air, the diﬀerence between the principal planes is equal to the nodal
space.
8
angle of a ray with respect to the lens axis. hence for u = 10 meters, and δ = 2
centimeters, the correction is u
= 9.96 meters. Even for close up photography,
this diﬀerence hardly matters. It only becomes important when δ is substantial
with respect to u — it should be noted that δ can be negative.
4 Pivot Points
Figure 4 shows the three planes and two lines from inﬁnity: the line IQ meeting
FP at Q and the line IP meeting SP at P. The distances from P and Q to LP
must of course be f. J and K are the lengths of the rays from the lens center,
L, to P and Q respectively. The points P and Q are pivot points, as will now
be discussed.
V
SP LP
FP
L
P
Q
f
f
f
f
J
K
φ
θ
θ
φ
γ
IP
IQ
Figure 4: Pivot points
Suppose LP is ﬁxed, and FP is moved parallel to itself, then SP must also
move in order for the lens equation, equation (1), to be satisﬁed. SP does not
move parallel to itself, as at ﬁrst might be thought, nor does it pivot about V.
It, in fact, turns about the pivot point P. P is the point where IP meets SP. By
deﬁnition, P must be f units from LP, and since there is only one such point on
the segment of IP below L, it follows that SP must pivot on P as FP is moved
parallel to itself.
A similar statement may be made about the reciprocal pivot point Q with
respect to movements of SP parallel to itself.
Later it will be seen that the near and far depth of ﬁeld limits are essentially
planes corresponding to movements of FP parallel to itself, hence they too must
9
pivot on P.
The tilt angle of the lens, φ, with respect to FP is easily seen to be given by
arcsin(f/J). Similarly the tilt with respect to SP is θ = arcsin(f/K), and γ,
the angle of SP with respect to FP, is φ +θ.
5 Determining the lens tilt
5.1 Using distances and angles
There are some authors, Merklinger(1993), who suggest that φ, the lens tilt an
gle, should be determined by guessing J. This can sometimes be done, since one
does not need to know φ very precisely — within a degree is usually satisfactory.
A more rigorous procedure is to measure the distance and angle of two points
in SP, and then calculate J.
Consider ﬁgure 5 where two rays of length d and D have been drawn from
the lens center on LP to SP. The longer ray is B degrees above the lens axis,
of length Z, from the lens center and the shorter one is β degrees below the
horizontal.
V
SP
LP
FP
P
J
γ
Z
y
x
X
Y
β
d
D
Β
K
Figure 5: J from two rays
10
For the ray above the lens axis, one has:
X
Y
=
Z
J
,
Y = DsinB,
X = Dcos B −Z,
J =
ZDsin B
Dcos B −Z
,
Z =
JDcos B
Dsin B +J
.
Although not needed here, it is useful to note that D = Z cos(γ)/ cos(B+γ).
A similar calculation may be made for the ray below the lens axis, but one
must adopt a convention about signs. There is a standard convention which
chooses B negative and β positive. It seems better in the present case to re
verse this convention; hence we will assume that angles below the lens axis are
negative. This leads to a common formula for both types of rays:
J =
Zδ sin α
δ cos α −Z
, (9)
Z =
Jδ cos α
δ sinα +J
. (10)
for an arbitrary ray of length δ and angle α.
If one can measure Z, then one may use equation (9) to ﬁnd J and hence
φ = arcsin(f/J).
If one cannot, or chooses not to measure Z, then one may eliminate Z by
using two rays and equating the values of equation (10) for each:
Jd cos β
d sin β +J
=
JDcos B
Dsin B +J
,
dD(sin Bcos β −cos Bsinβ) = J(d cos β −Dcos B),
dDsin(B −β) = J(d cos β −Dcos B),
1
J
=
d cos β −Dcos B
dDsin(B −β)
.
One may then ﬁnd φ, using φ = arcsin(f/J), from
sin φ =
f(Dcos B −d cos β)
dDsin(B −β)
. (11)
The angle, γ, SP makes with respect FP will be needed shortly. It is given
by γ = arctan(Z/J). If Z is known, then from equation (9), tanγ = (δ cos α −
Z)/δ sin α. If Z is not known, then one may ﬁnd Z by equating the values from
equation 9 for two rays, obtaining:
11
Z =
dDsin(B −β)
Dsin B −d sin β
,
and then,
tan γ =
Dcos B −d cos β
d sinβ −Dsin B
. (12)
The distance K shown in ﬁgure 5, is used in subsection 5.2. It is the conju
gate of Z for an untilted lens:
K =
f
Z −f
.
5.2 Using back focus
Figure 6 shows both a tilted lens plane (TLP), and a tilted ﬁlm plane (TFP).
Two points, A and C, on SP are shown in focus on TFP at a and c respectively.
SP
FP
TLP
LP
TFP
f
K
α
φ
J
c
a
b
α
A
C
γ
u
Figure 6: Back movements to obtain tilt
Note that the conjugates of SP points are on both TFP and on vertical
planes through the conjugates. Thus the distance from A to LP is conjugate to
the distance from LP to a, and these distances are the same both for a tilted
ﬁlm plane and for a vertical ﬁlm plane. Thus one may locate a and c by moving
FP. The angle α is given as the arctangent of (b − a)/(c − b), and it is easily
seen that φ may be found from
12
sin(φ) =
f
K
tan α = Qsin(α),
where
Q =
f
K cos(α)
=
1
(1 +m
) cos(α)
,
where K = uf/(u−f) = f(1 +m
) with m
the magniﬁcation for a ray through
the untilted lens center. The distance from the lens center to FP is of course
uf/(ucos(φ) −f) because ω in equation (2) is φ. See ﬁgure (20). In general Q
is not greatly diﬀerent from unity, hence one may often take φ = α.
Finally, one may ﬁnd γ from
tan(γ) =
Kf
K −f
1
J
=
f
K −f
tan(α) =
1
m
tan(α);
which leads to:
sin(φ) =
1
1
tan(α)
+
1
tan(γ)
.
In other words sin(φ) is half the harmonic mean of tan(α) and tan(γ).
5.3 Wheeler’s rules
A more practical expression for use in the ﬁeld is Wheeler’s rule of 60, given by
φ ≈ 60
(b −a)
(c −b)
(u −f)
u
,
where (u −f)/u is close to unity except for closeup work.
Since some cameras do not have angle scales, one can calculate the tilt in
terms of the distance ∆ that the top of the front standard moves. Let T be the
distance between the top of the front standard and the tilt axis, then Wheeler’s
tilt rule is
∆ = T
(b −a)
(c −b)
(u −f)
u
,
where (u −f)/u is close to unity except for closeup work.
Some cameras do not have movable backs, and must be focused by moving
the lens. The rules may still be used — but not for close up work.
EXAMPLE: A plane containing a nearby ﬂower and a distance
mountain sweeps from under the photographer’s feet passing through
both. The rail distance between the two points of focus on the cam
era is 12 mm. On the groundglass, the two images are about 60
13
mm apart. Wheeler’s rule of 60 indicates that the lens should be
tilted by 60*12/60=12 degrees. If the height of the front standard
measured from the tilt axis to the top of the standard is 25 cm, then
Wheeler’s tilt rule indicates that lens should be tilted forward until
the top of the standard has moved forward 25*12/60=5 cm.
5.4 Lens Movement
Figure (6) places the center of the lens at the same point for both a tilted and
untilted lens. But in fact the lens center is not at the same point for the two
situations. The lens moves as it is tilted. It moves slightly with center tilt
cameras, and substantially with base tilt cameras. This has little eﬀect on the
calculation of φ since if the lens center in ﬁgure (6) is moved horizontally by
∆, one has φ = arcsin(f/(J − θ)), where θ = ∆/ tan(γ), which is usually a
negligible change from φ = arcsin(f/J) because ∆ is seldom more than about
50 millimeters, and J is in meters.
Lens movement has more eﬀect on the focus position, since FP, for a tilted
lens, is at a horizontal distance of uf/(u cos(φ) −f) from the lens center, which
is larger than K = uf/(u −f) for the untilted lens. For a base tilt camera, the
FP for a tilted lens will be at approximately K − ∆, hence usually in front of
a. For a center tilt camera, the FP will usually be in back of a.
5.5 Back Tilts
Tilting the lens preserves perspective because the magniﬁcation deﬁned by v/u
remains unchanged. If h and y are distances as shown in ﬁgure (3), then
y/h = (y cos ω)/(hcos ω) = v/u for any tilt, and thus tilting leaves magniﬁ
cation unchanged. On the other hand, tilting the back changes the v/u ratio,
and the perspective. Parallel lines can be made to meet at vanishing points and
vice versa.
When the back is tilted, it is also necessary to tilt the lens to bring the
subject into focus. Figure (7) shows the relevant planes, where LP and FP are
the untilted lens and focal planes, and TB is the tilted back plane. The back is
tilted α degrees, and the desired lens tilt is given by α+β. It may be seen that
D =
J sin γ
sin(γ −α)
, (13)
and thus β is given by
β = arcsin
f sin(γ −α)
J sin γ
= arcsin
t
m + 1
sin(γ −α)
sin γ
,
where t = H/V in ﬁgure (7), and m is the magniﬁcation as in ﬁgure (6).
14
SP
FP
TLP
LP
f
J
γ
TB
D
TL
α
α
β
V
H
Figure 7: Lens tilt for tilted back
6 Depth of ﬁeld for parallel planes
Figures 8 and 9 show diagrams that will be used in deriving depth of ﬁeld, DOF,
equations for the case when SP, LP, and FP are parallel. The diameters of the
aperture and circle of confusion are d and c, respectively.
6.1 Near DOF limit
For ﬁgure 8, the circle of confusion arises because the image plane, IP is behind
FP. The near DOF distance is given by equation (3) as
Z
n
=
(v +x)f
(v +x) −f
.
By similar triangles, one sees that x = cv/(d−c), hence the near DOF limit
is
Z
n
=
vdf
vd +f(d −c)
=
uf
2
f
2
+ (u −f)Nc
=
u(H −f)
H +u −2f
, (14)
where N = f/d is the fnumber, and
H =
f
2
Nc
+f (15)
is the hyperfocal distance.
15
v
x
FP
LP
d/2
c/2
IP
Figure 8: DOF near
v
FP
LP
d/2
c/2
IP
y
Figure 9: DOF far
The Front DOF is u −Z
n
= u(u −f)/(H +u −2f).
If one takes H
a
=
f
2
Nc
, then one has
Z
n
=
uH
a
H
a
+ (u −f)
. (16)
One may also recast these equations in terms of the distance, t, from the focal
plane. Since t = u+v +δ, where δ is the nodal space, one has t = δ +u
2
/(u−f)
and u(t) = (t − δ)
1
2
+
1
4
−
f
t−δ
. It follows that the near distance from the
focal plane is
Z
nt
−δ =
u(t)f
2
f
2
+ (u(t) −f)Nc
+u(t)f/(u(t)−f) =
u(t)f
u(t) −f
u(t)f + (u(t) −f)Nc)
f
2
+ (u(t) −f)Nc
.
or perhaps more succinctly
16
Z
nt
−δ = Z
n
u(t)
(u(t) −f)
+
N
f
= Z
n
t
f(1 + 1/m)
+
Nc
f
= Z
n
t
f(1 + 1/m)
+
f
H
a
,
where m is magniﬁcation.
6.2 Far DOF limit
For ﬁgure 9, IP is in front of FP, giving
u
=
(v −y)f
(v −y) − f
.
By similar triangles, one has y = cv/(d +c), and
u
=
vdf
vd −f(d −c)
=
uf
2
f
2
−(u −f)Nc
.
When y is zero, the denominator vanishes because the rays from the object
are parallel, hence the far DOF limit is
Z
f
=
uf
2
f
2
−(u−f)Nc
=
u(H−f)
H−u
for u < H
∞ for u ≥ H
(17)
The Back DOF is Z
f
−u = u(u−f)/(H−u) for u < H and inﬁnity otherwise.
If one takes H
a
=
f
2
Nc
, then for (u −f) < H
a
one has
Z
f
=
uH
a
H
a
−(u −f)
. (18)
In terms of the distance, t, from the focal one has for u(t) < H
Z
ft
−δ =
u(t)f
2
f
2
−(u(t) −f)Nc
+u(t)f/(u(t)−f) =
u(t)f
u(t) −f
u(t)f −(u(t) −f)Nc)
f
2
−(u(t) −f)Nc
,
or perhaps more succinctly
Z
ft
−δ = Z
f
u(t)
(u(t) −f)
−
Nc
f
= Z
f
t
f(1 + 1/m)
−
Nc
f
= Z
f
t
f(1 + 1/m)
−
f
H
a
.
6.3 DOF
The DOF is the diﬀerence Z
f
−Z
n
, so
DOF =
2Nc(u−f)uf
2
f
4
−[Nc(u−f)]
2
=
2u(u−f)(H−f)
(H−f)
2
−(u−f)
2
for u < H
∞ for u ≥ H
, (19)
17
or
DOF =
2Ncvf
(v−f)
2
−(Nc)
2
=
2vf
3
(H−f)
(v−f)
2
(H−f)
2
−f
4
for v < Hm
∞ for v ≥ Hm
,
An interesting variant of this may be obtained by substituting equation (6)
into equation (19):
DOF =
2Nc
m+1
m
2
1−K
for K < 1,
∞ for K ≥ 1
(20)
where K = (
Nc
mf
)
2
, which for small K (as might occur in macro photography)
becomes
DOF ≈ 2Nc
m+ 1
m
2
= 2Nc
S
I
+
S
I
2
, (21)
where S is the subject height, and I the image height.
In general, one should use
DOF =
2Nc(m+ 1)
m
2
−(
Nc
f
)
2
. (22)
This equation shows how DOF decreases as f increases. It is interesting to note
that the denominator of equation 22 vanishes when u equals the hyperfocal
distance as given by equation 15. In practice, this means that DOF depends
on f when u > H/2, say, but not for closer subjects.
The near and far limits in terms of m are:
Z
n
=
f(1 +m)
m +
Nc
f
, (23)
and
Z
f
=
f(1+m)
m−
Nc
f
for mf > Nc
∞ for mf ≤ Nc
(24)
6.4 Circles of confusion
The human eye has a resolution limit of about 5 line pairs per millimeter (l/mm)
at the nearest distance of distinct vision (about 25 cm for a normal eye). This
means that a normal human eye can just discriminate between ﬁnely drawn
lines separated by 0.2 mm. One can check this for oneself with a lens chart. It
follows that 5 l/mm is the maximum resolution needed by a print. The ﬁeld
of view of human vision is about 60 degrees, which corresponds to about 290
mm at 25 cm. The diagonal of an 8x10 print is 325 mm, and viewing such a
print at 25 cm completely ﬁlls the ﬁeld of view. For such a print viewed at such
a distance, the perspective will be correct if the photograph was taken with
18
a standard lens
3
. This means that to preserve perspective, one should view a
16/20 print at 50 cm, and other prints proportionally.
For an 8x10 contact print, one should have at least 5 l/mm. When the image
size is smaller, magniﬁcation comes into play: thus there will be a factor of 8
for a 35 mm image. This implies geometrically that the ﬁlm resolution should
be eight times 5 l/mm or 40 l/mm. The reciprocal of this, 0.025 mm, gives
the spacing between lines, and the diameter of a circle at the limit of human
discrimination. This circle is called the “circle of confusion.”
The circle of confusion thus depends on the degree of magniﬁcation of the
image with respect to the size of a print taken with a standard lens. As may
be seen throughout these notes, the diameter of this circle, c, is an important
variable. For DOF, setting it properly means that those objects at the DOF
limits will just be resolvable in a print. For a 35 mm camera, it should be at
most about 0.025/mm, and should decrease as the print size increases past 8x10.
6.5 DOF and format
There is always heated argument about the DOF achieved from various formats.
The general view is that larger formats have smaller DOF. To a large extent this
opinion derives from a selective modiﬁcation of the parameters: leaving some
ﬁxed and changing others.
If one ﬁxes the fnumber, then the DOF decreases as the format increases,
but this is not so if the fnumber is scaled as the other parameters are scaled.
Figure (10) on the left shows the DOF when N = 16, c = 0.025 mm, m = 0.01,
and f = 50 mm. The x axis indicates the format in terms of the size of the
short side of the ﬁlm — 1 for 35 mm, 4 for 4x5. The lens to subject distance,
u, is approximately constant at 5 m for the x’s shown. One concludes from this
that DOF decreases with format size when the fnumber is ﬁxed, but remains
unchanged or slightly increases when the fnumber number is scaled with the
other parameters.
For macro photography, with m = 1 and the other parameters as above,
Figure (10) on the right shows the DOF in the ﬁxed and scaled cases — note
that the DOF scales are in millimeters on this graph. One has u = 100 mm for
the 35 mm format and u = 400 mm for the 4x5 format.
6.6 The DOF equation
An interesting equation similar to the Gaussian lens equation may be derived
for the near and far DOF limits.
From ﬁgures 8 and 9 one sees
d
c
=
v +x
x
=
v −y
y
= r,
3
One who’s diameter is approximately equal to its focal length
19
Figure 10: DOF for Fixed and scaled N
thus
1
v +x
=
1
v
r −1
r
,
1
v −y
=
1
v
r + 1
r
,
and since
1
Z
n
+
1
v +x
=
1
u
+
1
v
,
1
Z
f
+
1
v −y
=
1
u
+
1
v
,
one has by addition
1
Z
n
+
1
Z
f
=
2
u
+
2
v
−
1
v
r −1
r
+
r + 1
r
,
or the DOF equation
1
Z
n
+
1
Z
f
=
2
u
. (25)
A more straightforward derivation is to sum the reciprocals of equations (14)
and (17).
6.7 Hyperfocal distance
The hyperfocal distance in equation (15) is the distance at which the far limit
of DOF becomes inﬁnite. Substituting this as u into the formula for the near
limit, equation (14) gives
20
Z
n
=
H
2
,
thus focusing at the hyperfocal distance puts everything in focus from half this
distance to inﬁnity. It is worth noting that H =
f
2
Nc
+f = f(1 +
f
Nc
).
Another often quoted deﬁnition for “hyperfocal distance” is the value of Z
n
when u = ∞. As u increases, Z
n
in equation (14) may be seen to approach
f
2
/(Nc), a good approximation to H.
As a practical matter, the hyperfocal focus is best determined from distances
on the image side of the lens. Simple substitution into equation (15) shows
v
H
= f +Nc = f(1 +
Nc
f
), (26)
where v
H
is the position of the standard when focused on the hyperfocal dis
tance. Thus one may focus on inﬁnity, and then move the standard by Nc to
obtain focus at the hyperfocal distance.
6.8 Approximations
These equations simplify by noting that in many cases u − f ≈ u, giving the
usually cited formulas
H
≈
f
2
Nc
. (27)
Z
n
=
uf
2
f
2
+uNc
=
Hu
H +u
, (28)
Z
f
=
uf
2
f
2
−uNc
=
H
u
H
−u
for u < H
∞ for u ≥ H
(29)
DOF
=
2Ncu
2
f
2
f
4
−(Ncu)
2
=
2H
u
2
H
2
−u
2
for u < H
∞ for u ≥ H
(30)
The approximate DOF is slightly larger than the exact DOF. A calculation
shows that Z
n
(1 − Nc/f) ≈ Z
n
, and Z
f
(1 − Nc/f) ≈ Z
f
, and that DOF
≈
DOF + (Z
n
+ Z
f
)Nc/f. It is hard to imagine situations in which these small
diﬀerences will matter.
6.9 Focus given near and far limits
6.9.1 Object distances
An interesting question, is “what is the focus distance, given the near and far
DOF limits?”
21
The DOF Equation, equation(25), is
1
Z
n
+
1
Z
f
=
2
u
,
which when solved for u gives
u =
2Z
n
Z
f
Z
n
+Z
f
=
2αx
1 +α
, (31)
where x = Z
n
and α = Z
f
/Z
n
. Note also that u is the product of the limits
divided by their average, and that the commonly stated one third rule
4
is correct,
nontrivially, only when α = 2; although many will be satisﬁed when using it in
the range of, say, 2 ±
1
2
.
One may solve for N given f in equation (14), obtaining
N =
f
2
c
Z
f
−Z
n
(2Z
f
Z
n
−f(Z
f
+Z
n
))
=
f
2
c
α −1
(α(2x −f) −f)
, (32)
and for f given N:
f =
1
2
Nc
Z
f
+Z
n
Z
f
−Z
n
−1 +
1 +
8Z
f
Z
n
Nc
Z
f
− Z
n
(Z
f
+Z
n
)
2
,
=
1
2
Nc
α + 1
α −1
−1 +
1 +
8αx
Nc
α −1
(α + 1)
2
.
Thus from two distances x and αx, with α > 1 one may obtain the focus
distance and either the fnumber or the focal length that will make these two
distances the near and far DOF limits. It is useful to note that equation (31)
ranges between 1x and 2x, and that 2x, for x > 0, corresponds to focus at the
hyperfocal value. It is also interesting that approximations for f and N may be
obtained by equating the focus distance to the hyperfocal distance.
6.9.2 Image distances
The expressions simplify considerably when one uses the image distances z
n
and
z
f
corresponding to Z
n
and Z
f
respectively. For u one has
1
u
=
1
f
−
1
θ
, (33)
where θ is the harmonic mean of z
n
and z
f
:
θ =
2z
n
z
f
z
n
+z
f
.
Equation (33) may also be obtained from equation (2).
4
“The focus point is one third of the distance between the near and far limits.”
22
Since θ = λ/µ, where λ is the square of the geometric mean and µ the
arithmetic mean of z
n
and z
f
, and since all three means diﬀer by at most about
1%, one can write equation (33) as
1
u
≈
1
f
−
1
µ
.
For N given f one may write from equation (32)
cN = ρ
f
µ
=
ρ
1 +m
, (34)
where ρ = (z
n
−z
f
)/2, and m is magniﬁcation.
When the magniﬁcation is small, one has the approximation
cN ≈ ρ,
which is likely that used by Sinar for their DOF calculation.
The expression for f given N is of course
f =
µcN
ρ
.
7 Depth of ﬁeld, depth of focus
The function relating DOF and ρ may be obtained by assuming the depth of
focus range 2ρ is equally divided by the focus point v. Thus (v −ρ) and (v +ρ)
are the conjugates of Z
f
and Z
n
. This gives
DOF = Z
f
−Z
n
=
2f
2
ρ
(v −f)
2
−ρ
2
=
2f
2
ρ
(mf)
2
−ρ
2
, (35)
from which one may express ρ as a function of DOF and m:
ρ =
f
2
DOF
¸
−1 +
1 +
mDOF
f
2
¸
. (36)
Equation (34) may be used to ﬁnd m = (ρ −Nc)/Nc, which then gives ρ as
a function of DOF:
ρ =
Ncf
2
f
2
−(Nc)
2
¸
1 +
Nc
DOF
+
1 +
Nc
DOF
2
−
1 −
Nc
f
2
¸
,
or
ρ =
Ncf
2
f
2
−(Nc)
2
¸
1 +
Nc
DOF
+
2
Nc
DOF
+
Nc
DOF
2
+
Nc
f
2
¸
.
23
8 Fuzzy images
Points image as circles on the ﬁlm. Some of the image circles are so small that
the eye cannot distinguish them on a print or slide. These are those between
the DOF limits. Points in other regions, image as circles which the eye can
distinguish, and objects made up of such circles are fuzzy. Knowing the degree
of fuzziness for an object can be useful; as when one wants letters on a sign
in the background to be unreadable. This may be done by calculating the size
of an object at a given distance, which will produce an image with diameter k
times the diameter of a circle corresponding to a point image at that distance.
Thus an object with k = 2 will be quite fuzzy and impossible to recognize, while
one with k = 10 will probably be recognizable, although still fuzzy.
Figure 11 shows the three planes as well as a plane, P, and its image plane,
IP with a circular
5
object of diameter S on P. The image of this object has
diameter r on IP, and t on FP. The diameter of the aperture is d The diameter
of the image of an object point is shown as p. If P were the near DOF plane,
then p would equal c. The problem is to calculate S corresponding to a multiple,
say k, of p.
SP
P
LP FP
IP
r
t d S
u
u'
v
x
p
Figure 11: Object in front of SP
The magniﬁcation of the object with respect to IP is r/S = (v +x)/u
, thus
r = S
v +x
u
= S
f
u
−f
,
from equation (3).
From similar triangles (d −r)/(t −r) = (v +x)/x one ﬁnds
t =
rv +xd
v +x
.
5
A ﬂat disk without depth.
24
Again using equation (3), one ﬁnds
x = v +x −v =
u
f
u
−f
−
uf
u −f
=
f
2
(u −u
)
(u
−f)(u −f)
. (37)
From similar triangles p/d = x/(v +x), and one obtains
p =
xd
v +x
.
Let t be a multiple of p, say t = kp, then
t =
rv +xd
v +x
= kp = k
xd
v +x
,
thus
(k −1)xd = rv = r
uf
u −f
= S
f
u
−f
uf
u −f
,
and using equation (37),
S = (k−1)d
(u
−f)(u −f)
f
2
u
x = (k−1)
f
N
u −u
u
= (k−1)
xf
2
Nv(v +x −f)
, (38)
with N = f/d.
At the near DOF limit, this becomes
S = (k −1)
f
N
1 −
Z
n
u
.
It is worth noting that the above derivation holds if S is not symmetrical
about the lens axis.
Figure 12 shows the planes with P to the left of SP. The argument is essen
tially the same as above. The only small problem is ﬁnding the similar triangles
involving t. The inset shows the triangles, which give
d +t
t −r
=
v
y
,
thus,
t =
rv +dy
v −y
.
The equation is
S = (k −1)
f
N
u
−u
u
= (k −1)
yf
2
Nv(f +y −v)
, (39)
and when P is the far DOF limit:
25
LP
d
u
u'
SP
IP
FP
p
v
r
y
t
P
S
(d+t)/2
(tr)/2
LP
IP FP
Figure 12: Object behind SP
S = (k −1)
f
N
Z
f
u
−1
.
The derivation does not depend on the symmetry of S about the lens axis.
Useful expressions for t and S in terms of each other are:
t =
fu −u
d
u
(u −f)
+
fu
u
(u −f)
S,
S =
u
(u −f)
fu
t −
u −u
d
u
.
9 Eﬀects of diﬀraction on DOF
9.1 Theory
Light from a point source, diﬀracted by a circular opening, is focused by a
lens not as a geometrical point, but as a disk of ﬁnite radius surrounded by
dark and bright rings. This is due to the fact that the path of light near
an edge is longer than that away from the edge causing a phase diﬀerence,
and complete interference occurs at a certain distance from the center, and at
harmonic distances further away.
From a study of the appearance of the diﬀraction patterns of closely spaced
points, Lord Rayleigh concluded that two equally bright point sources could
just be resolved by an optical system if the central maximum of the diﬀraction
pattern of one source coincided with the ﬁrst minimum of the other. The disk
26
deﬁned by this ﬁrst minimum is called an Airy disk. The angle α from the center
of the lens subtended by two points with this spacing is given by sin(α) =
(1.22λ)/(nD), where λ is the wavelength of light, D is the diameter of the
aperture, and n the index of refraction. For an object at distance u, one ﬁnds
the object distance to be s = sin(α)u. Since u/D = f(1 +
1
m
)/D = N(1 +
1
m
),
where f is focal length, m is magniﬁcation, and N is the fnumber, one has:
s = 1.22N(1 +
1
m
)λ/n ≈
N(1 +
1
m
)
K
λ
n
mm, (40)
Since visible light ranges from about 400 × 10
−
6 mm to about 700 × 10
−
6
mm, K
λ
varies from about 1000
1
(mm)
to about 2000
1
(mm)
. Often one takes
K
λ
= 1500
1
(mm)
for a medium wavelength λ = 550 × 10
−
6 mm: see Sears
(1958), p260.
The distance s
on the other side of the lens is given by s
= ms(n/n
), where
n
is the index of refraction on the image side of the lens. This is
s
≈
N(1 +m)
K
λ
n
mm, (41)
The spacial frequency 1/s gives lines per millimeter, l/mm, as the frequency
of two just separable lines. The diameter of the diﬀraction circle of confusion,
c, is s. Test charts allow the selection of “just separable” pairs of lines, called
either “line pairs per millimeter”, lp/mm, or simply “lines per millimeter”,
l/mm. These correspond to the spacial frequency 1/s. For test charts, black
bars are used separated by spaces of the same width, thus a spacial cycle consists
of a black bar and a space, with s/2 the common width.
It is generally agreed that normal human vision can just resolve about 5
l/mm on a test chart at the minimum distance of distinct vision (about 250mm).
This may be conﬁrmed by inspecting an Edmund USAF pattern. It follows that
for human vision, s ≈ 1/5 = 0.2 mm or 0.008 in. Assuming the human eyeball
is about 25 mm, this gives m = 0.1 which when substituted in equation (40)
gives N ≈ 25 mm. Since N = f/D, and f ≈ 25 mm, this gives D ≈ 1mm,
which is a bit small, but in the ball park (In bright light the iris can close down
to about 2 mm.).
9.2 Data
Diﬀraction aﬀects resolution so that the parameter c used in equations, such as
equation (34), changes as other parameters change. Table 1
6
shows resolution
in l/mm as a function of N and depth of focus, δ mm
7
, about the point of focus.
The 35 observations in this table were obtained by using a 150 mm Sironar lens
with Velvia. The contrast between the dark lines and the white background was
approximately 5 stops, placing the contrast ratio, 35, approximately midway
6
records 153157
7
Note: δ = 2ρ in equation (34)
27
δ mm
fnumber 0 2.5 5 7.5 10
64 23 23 20 20 20
45 23 23 26 23 23
32 26 32 23 16 10
22 32 26 14 9 6.4
16 32 18 9 6.4 4.5
11 23 18 6.4 4.5 3.2
8 26 9 4.5 2.5 2.25
Table 1: Raw resolution data for Velvia
δ mm
fnumber 0 2.5 5 7.5 10
64 22 20 20 20 17
45
32 28 25 20 17 11
22 28 22 14 8 9
16
11 31 11 6.1 3.9 3.1
8 28 4.3 3.1 2.2 1.7
Table 2: Raw resolution data for TMax 100
between the usual 1.6 and 100 contrast ratios. The USAF target was 1500 mm
distant from the lens giving a magniﬁcation of 1/9.
The data was smoothed by a low order polynomial and is shown as Figure
13. The residuals had a sample standard deviation of 3.1 l/mm; which is of the
order of experimental error, since one can read the USAF charts only to the
nearest 2 l/mm. Table 2
8
shows a set of data taken with TMAX 100 using a
210mm Sironar lens — all other parameters are the same as for Velvia.
In Section 9.6 it will be seen that the data for δ = 0 does not agree with
equation (43). This suggests that the data may be in error. It seems likely
that the focus when taking the data was not at δ = 0, but at some small value,
perhaps δ = 0.5 or larger.
To preserve perspective, one should view a picture at a distance consistent
with that of the taking lens. For normal lens’s where the focal length is approx
imately equal to the ﬁlm size, this means that one views a picture x inches high
from x inches. For a 4x5 ﬁlm, this means that an 8x10 will be at the distance
of distinct vision, and so the diameter of the onﬁlm circle of confusion should
be one half s or 0.1 mm, corresponding to a spacial frequency of 10 l/mm.
The locus of equation (34) is marked on ﬁgure 13 by the line labeled c = 0.1.
Lines for c = 0.05 and c = 0.2 are also shown. The c = 0.1 lines seems to be
8
record 149
28
8
16
32
64
F

S
T
O
P
2
0
2
2
2
4
2
4
2
8
1
8
4
6
1
0
1
2
1
6
c=0.05
c=0.1
c=0.2
0
2 4 6 8 10
δ mm
Figure 13: Resolution as a function of N and δ for m=1/9
somewhat above the 10 l/mm contour, but the data has a standard deviation
of at least 2 l/mm. Since 4x5 prints are often printed at 16x20, one would need
a ﬁlm resolution of 20 l/mm, which corresponds to c = 0.05. The line c = 0.05
shown in ﬁgure 13 is slightly above the 20 l/mm contour. Both the data and the
choice of ﬁtting model could be improved upon. The results are good enough,
however, to show that equation (34) is reasonable.
As an additional check, an array of dollar bills was photographed and the
ﬁlm examined under magniﬁcation. In ﬁgure 17 the dollars were arranged 30
mm apart on a slanting board. A Pentax 645 camera with a 150 mm lens at f32
was positioned 1500 mm from the central pair of bills (numbers 8 and 9, marked
with dots on their corners). The DOF using c = 0.05 is 290 mm, with the front
and back DOF being 130 mm and 160 mm. This ranges from bill number 5
through bill number 12. The magniﬁcation is 1/9 = 150/(1500 − 150), hence
a 9 power loupe will bring the image to full size. An 8 power loupe was used.
Noticeable degredation seems to occur in the neighborhood of bills 5 and 12, so
the choice of c = 0.05 for a 645 format seems reasonable.
9.3 Resolution
It is useful to observe how the contours change in ﬁgure 13. For large δ, the
lines per millimeter increase steadily with N because of the shrinking aperture.
For small δ’s, however, N increases to a maximum and then decreases. This is
due to the fact that lines per millimeter, increases directly with N for depth of
focus, and inversely for diﬀraction. Figure 15 indicates the eﬀect of decreasing
29
8
16
32
64
F

S
T
O
P
2
0
2
2
2
4
2
4
2
8
1
8
4
6
1
0
1
2
1
6
Hansma
0
2 4 6 8 10
δ
Opt Opt2
mm
Figure 14: Optimum N as a function of δ for m=1/9
the aperture on the image circle. This ﬁgure may be compared with ﬁgures
8 and 9. For a ﬁxed depth of focus with a constant distance between IP and
FP, decreasing the aperture is seen to decrease the image circle, thus increas
ing the resolution. The eﬀect of diﬀraction on resolution is just the opposite,
since decreasing the aperture makes the image more diﬀuse and decreases the
resolution.
For depth of focus, equation (34) gives
L
δ
= (2N(1 +m))/δ.
From equation (41)
9
for diﬀraction, one has
L
d
= K
λ
/[N(1 +m))].
The point at which the diameter of the depth of focus circle of confusion
equals the diameter of the diﬀraction circle of confusion is obtained by equating
L
δ
to L
d
, giving
N =
K
λ
δ/2
1 +m
, (42)
which is the maximum N for a given δ, since the product L
d
L
δ
= 2K
λ
/δ does
not depend on N.
9
Equation (40) is not appropriate since here one needs to compare resolutions on the image.
30
v
x
FP
LP
d/2
c/2
IP
Figure 15: Eﬀect of aperture on the image circle.
Hansma (1996) attempted to combine depth of focus and diﬀraction. He
chose an empirical combination of depth of focus and diﬀraction by taking the
RMS of the two. He seems to have made several errors such as using 2s for
the diﬀraction circle of confusion, ignoring the magniﬁcation factor in equation
(40), and even confusing equations (40) and (41). In any case, Hansma found the
optimum N to be N =
(375δ) for m = 0. This is 1/
(2) of equation (42). The
particular choice of RMS has no eﬀect, since any linear combination of powers
of the two circle of confusion expressions will produce the same optimum.
Equation (42) seems to agree better with the observed optimum than does
Hansma’s. Equation (42) is marked “Opt” in ﬁgure (14). The curve “Opt2”
represents 2L
δ
= L
d
, which may help to visualize the eﬀect of diﬀraction on
resolution. The apparent intersection of the contours with the δ = 0 vertical
are likely to be an artifact the polynomial form and measurement error.
9.4 Format considerations
It is worth noting that equation (42) does not depend on format. Although the
resolution values will diﬀer among formats, the relationship between N and δ
does not change. Equation (36) gives ρ = δ/2 as a function of magniﬁcation, m,
DOF, and f. The limit of equation (36) as DOF goes to inﬁnity is δ = 2mf.
This assumes depth of focus on both sides of the focus. When the far limit
is inﬁnite, the near limit is half the hyperfocal distance, and this works out
to δ ≈ 2mf where m is the magniﬁcation for an object at half the hyperfocal
distance. For a ﬁxed object size at a ﬁxed distance from the lens, the m and f
for 35 mm are both about 1/3 that for 4x5, reducing mf by about 1/9. It is
important to note that these changes do not eﬀect DOF, as may be seen from
equation (35) by substituting f/3, m/3 and ρ/9.
Figure (16) shows equation (42) for this 35 mm scaling.
The diﬀraction resolution L
d
is aﬀected in a very interesting way by format.
From Section 6.5, N must change as format changes in order to keep DOF
31
8
16
32
64
F

S
T
O
P
Opt
Opt2
0
0.2 0.4 0.6 0.8
1.0
c=0.025
δ mm
Figure 16: Optimum N as a function of δ for 35 mm
constant: but L
d
is inversely proportional to N(1 + m), hence smaller formats
must have greater resolution. The change from 4x5 to 35 mm is about 1/3,
hence 35 mm images must produce about 3 times the resolution of 4x5 images,
and they do!
Another way to view this is to examine equation (41) with N = f/A, where
A is the aperture. For any format, the smallest aperture that is practical is
about 2 mm, hence resolution is a function of the distance between lens and
ﬁlm (f(1 +m)). Since this is smaller for smaller formats, their resolutions must
be greater.
9.5 Minimum aperture
The curve from equation (34) using the appropriate c lies well below the Opt
curve in ﬁgures (14) and (16). The δ at which the Opt curve meets the c curve
is 2K
λ
× c
2
, which for K
λ
= 1500, is about 1.8 for 35 mm and 30 for 4x5 —
well outside any practical range. For any format the point (N, δ) is usually well
away from the optimum curve, except at extreme depth of focus limits for the
format. At such extremes the point (N, δ) comes close to the optimum, and any
substantial increase in N will decrease resolution. Lens manufacturers use this
to choose the minimum fnumber for their lenses.
32
Figure 17: DOF determination
9.6 Theoretical curves
One must combine L
d
and L
δ
in order to ﬁnd the actual resolution. The com
bination should approach one of the resolutions as the other becomes extreme.
One choice is (1/L
α
)
α
= (1/L
δ
)
α
+ (1/L
d
)
α
, which gives
L
α
=
K
λ
N(1 +m)
((K
λ
δ/2)
α
+ (N(1 +m))
2α
)
1/α
, (43)
Substituting equation (42) into this gives the maximum L
α
= 2
−1/α
Kλ
δ/2
.
The agreement between equation (43) and the data is good, suggesting that it
captures a fundamental aspect of the problem. Values calculated from L
α
agree
reasonably well with ﬁgure (14) in view of the 2 l/mm error in the data, except
for δ = 0. Table 3 shows values calculated from equation (43), which may be
compared with tables 1 and 2. A contour plot of this data is shown in ﬁgure
(18), and may be compared with ﬁgure (13). The data at δ = 0 was not used
to make the contour plot, thus the plot does not show the singularity at δ = 0.
The predictions for the 35 mm case produce resolutions of a magnitude sim
33
ilar to what one expects from 35 mm. The implication of this is that diﬀerences
in the formats are principally the result of size and not of superior lens quality
for smaller formats! Both lens quality and focus precision play a role, but the
higher resolutions for small format appear to be due to operating in the range
of smaller δ’s. This is consistent with the conclusion in Section 9.4.
δ mm
fnumber 0 1 2 4 6 8 10
64 21.1 20.9 20.2 19.4 15.8 13.6 11.8
45 30.0 28.7 25.7 19.2 14,6 11.5 9.5
32 42.2 36.3 27.2 16.4 11.4 8.7 7.0
22 61.4 38.2 22.7 12.0 8.1 6.1 4.9
16 84.4 32.8 17.4 8.8 5.9 4.4 3.6
11 122.7 24.0 12.2 6.1 4.1 3.2 2.4
8 168.8 17.7 8.9 4.4 3.0 2.2 1.8
Table 3: Theoretical resolution from eq (43) with α = 2.
3
5
1
0
1
5
2
0
2
5 3
0 3
5
8
16
32
64
δ mm
F

S
T
O
P
0 2 4 6 8 10
Figure 18: Contours for theoretical resolution from eq (43) with α = 2.
34
10 Depth of ﬁeld for a tilted lens
10.1 Near and far DOF equations
The DOF equations above hold for tilted lenses with minor modiﬁcations. Fig
ure 19 shows the three planes SP, LP, and FP with a ray at angle ρ. Planes F
and I are parallel to LP and have been drawn through the intersections of the
ray with SP and FP. The distances of F and I from LP are u and v, respectively.
Equation (14) may thus be used to ﬁnd the near DOF which is labeled u
in
ﬁgure 19. Let h be the distance from the lens center L along the ray to SP, then
u = hcos(φ +ρ), thus equation (14) becomes
u
=
hf
2
cos(φ +ρ)
f
2
+ [hcos(φ +ρ) −f]Nc
,
and the near DOF along the ray is Z
n
= u
/ cos(φ +ρ).
There is one minor problem: the diameter of the circle of confusion is not
correct. The above assumes the image is on I not FP. The image on FP is an
ellipse, and its maximum diameter varies slightly with ρ; thus the near DOF for
a ray at angle ρ is:
Z
n
(ρ) =
hf
2
f
2
+ [hcos(φ +ρ) −f]g
. (44)
where g ≈ Nc.
The graph of this equation, with constant g, is a straight line, shown as
NP in ﬁgure 19. It corresponds to the image plane IP behind FP, and passes
through the pivot point.
Similar calculations follow through when the ray is below the lens axis,
and following the sign convention used above, result in equation (44) with ρ a
negative quantity.
The far DOF limit is derived in the same way and is
Z
f
(ρ) ≈
hf
2
f
2
−[hcos(φ+ρ)−f]g
for h < H(ρ),
∞ for h ≥ H(ρ)
(45)
where H(ρ) = f
2
/(g cos(φ +ρ)).
10.2 Near and far DOF equations in terms of ρ
It is often more useful to calculate these limits in terms of ρ instead of h. To
this end, it may be seen from ﬁgure (20) that
hcos ρ
tan γ
= J +hsin ρ,
hence
35
SP
LP
FP
IP
Z
n
I
u
h
NP
ρ
φ
ρ + φ
L
v
F
u'
v'
VP
Figure 19: Tilted DOF diagram
h =
J sin γ
cos(ρ +γ)
,
and thus, using J = f/ sin φ,
Z
n
(ρ) =
f
2
sin γ
f sin φcos(ρ +γ) +g cos ρ sin(γ +φ)
, (46)
and
Z
f
(ρ) =
f
2
sin γ
f sin φcos(ρ+γ)−g cos ρ sin(γ+φ)
for nonnegative denominator
∞ for negative denominator
(47)
36
SP
LP
FP
h
ρ
L
J
a
b
Figure 20: Finding h from ρ
10.3 Focus given near and far limits
10.3.1 Object distances
Just as for parallel planes, one may obtain the focus distance from the near and
far limits. An equation similar to the DOF Equation, equation (25), holds for
Z
n
(ρ) and Z
f
(ρ);
1
Z
n
(ρ)
+
1
Z
f
(ρ)
=
2
h
, (48)
which is most easily seen by assuming g constant and summing the reciprocals
of equations (44) and (45).
The DOF for a given ray angle ρ is
DOF(ρ) =
2g(hcos(φ +ρ) − f)hf
2
f
4
−[g(hcos(φ +ρ) −f]
2
.
An expression for N given f is
37
N =
f
2
c
Z
f
(ρ) −Z
n
(ρ)
2Z
n
(ρ)Z
f
(ρ) cos(φ +ρ) −f(Z
n
(ρ) +Z
f
(ρ))
, (49)
and for f given N:
f =
g
2Z
n
(ρ)Z
f
(ρ) cos(φ +ρ) −fZ
n
(ρ) +Z
f
(ρ))
Z
f
(ρ) −Z
n
(ρ)
.
One might use these expressions in calculating the lens tilt by replacing the
distance to SP along one of the rays with the distances of the desired near and
far limits.
10.3.2 Image distances
Using equation (2) in the form hf/(hcos(φ +ρ) −f) in equation (49) gives
cN = δ
f
µ
,
where δ = (z
n
(ρ) − z
f
(ρ))/2 and µ = (z
n
(ρ) + z
f
(ρ))/2, with z
n
(ρ) and z
f
(ρ)
the conjugates of Z
n
(ρ) and Z
f
(ρ) via equation (2). Since the parameters c, N,
and f are constant in this formula, the ratio δ/µ is also constant for all ρ, and
one may choose to measure δ and µ along the original lens axis, with ρ = 0.
In this case, one has µcos(φ) = f(1 + m), and remembering that m is a
constant for all angles, one has
cN =
δ cos(φ)
1 +m
. (50)
This may be compared with equation (32) in section 6.9.2, which is equivalent
for practical purposes, since φ is never very large.
The expression for f given N is of course
f =
µcN
2δ cos(φ)
.
10.4 The graph of the DOF limits
The tangent of the graph of Z
n
(ρ) is
T = tan ρ +
J
Z
n
(ρ) cos ρ
.
On substituting Z
n
(ρ) as given by equation (46) into this, and using J =
f/ sin φ, one has
T =
1
tan γ
+
g
f
1
tan φ
+
1
tan γ
,
38
which shows that the graph depends on ρ only through g, and hence if g is
constant, the graph is a straight line. Similar comments apply to Z
f
(ρ): the
equation is the same, save for a negative sign in front of g.
An interesting observation is that the rise of the Z
n
(ρ) graph above SP at
the hyperfocal distance is approximately equal to J. Let H = f
2
/g, and note
that H/ tan γ is on SP, then one sees from T, above that the rise above SP is
J cos φ +
f
tan γ
.
This is practically diﬀerent from J, only for very small γ or large φ. The fall
for Z
f
(ρ) is also approximately equal to J.
Since arctan(T) is
π
2
−γ
+
, where γ
+
is the angle of the near DOF plane with
respect to the vertical, one has
1
tan(γ+)
=
1
tan γ
+
g
f
1
tan φ
+
1
tan γ
, and
1
tan(γ−)
=
1
tan γ
−
g
f
1
tan φ
+
1
tan γ
,
with γ
−
the angle of the far DOF plane with respect to the vertical.
10.5 Variation of maximum diameter of the ellipse of con
fusion with ρ
We will show that the variation of the maximum diameter of the ellipse of
confusion is small, and consequently the DOF limits are approximately linear.
Figure 21 is a modiﬁed portion of ﬁgure 19. The aperture is the slightly
darkened portion of LP with diameter d. The corresponding image on I, cen
tered
10
about the slanted ray, has diameter
11
δ. On FP, the image is an ellipse
with extents c1 above and c2 below the ray. We desire c = c1+c2 to be constant,
and thus must ﬁnd δ.
The distance, j, along the ray from LP to FP is v/ cos(ρ + φ), and similar
triangles give δ/x = d/(j +x), or
x =
δv
(d −δ) cos(ρ +φ)
,
or after some manipulation:
x =
δ
d −δ
f sin γ
cos ρ sin(γ −φ)
.
Let p =
δ
2
cos φ, then one may write
10
LP and I are parallel, hence the ray divides δ into two equal parts. Even more, the size
of δ is constant because x/j = δ/(d −δ) is constant — circles of confusion remain circles and
do not turn into ellipses.
11
Note: g = Nδ.
39
L
LP
FP
IP
I
v
ρ
φ
ρ + φ
x
c1
c2
b
a
p
p
e
e
δ/2
δ/2
j
d/2
d/2
φ
ρ
Figure 21: Ellipse of confusion diameter for tilted lenses
c1 = p −e
p +a
b +e
,
c2 = p +e
p −a
b −e
,
where e =
δ
2
sin φ, a = xsin ρ, and b = xcos ρ.
Adding and simplifying gives
c = δ cos φ +δ sin φ
(
δ
2
)
2
cos φsin φ −x
2
sin ρ cos ρ
(xcos ρ)
2
−(
δ
2
sin ρ)
2
¸
(51)
and for paraxial rays, the expression on the right diﬀers negligibly from δ; hence
one may reasonably take g = Nc as a constant in equations (46) and (47). Figure
22 shows a comparison of the near DOF plane, with a graph of equation (46)
using g = Nδ obtained by solving equation (51) for δ as a function of c. The
parameters are φ = 5
◦
, γ = 80
◦
, F = 150mm, and N = 32. The dotted curve,
representing the graph of equation (46), becomes inﬁnite when ρ ≈ 22.6
◦
: the
40
angle of the near DOF plane. This is an extreme example; the diﬀerences are
usually smaller.
SP
LP
FP
L
1
5
º

4
0
º

6
5
º

8
5
º
0º
5º
10º
NP
7.5º
0mm 500mm 1000mm 1500mm 2000mm
Figure 22: Comparison of near DOF plane (solid) with true curve (dotted).
11 Image control
11.1 Exposure
There are two quantities of prime interest: the luminance, L, of the scene and the
exposure, H
f
, at the focal plane. The luminance is modiﬁed into the exposure
by the camera controls, which are the fnumber, N and the exposure time t.
The exposure is decreased by increasing N, and increased by increasing t. Thus
H
f
is made smaller by increasing the ratio N
2
/t. This means that if one ﬁxes
the scene luminance, then the ratio L/H
f
increases with N
2
/t. In fact, L/H
f
is proportional to N
2
/t, with the constant of proportionality being ﬁxed by the
scene and the ﬁlm density desired.
Since N and t are under the control of the photographer, and the luminance
is ﬁxed by the scene, the photographer needs only two things to make an ap
propriate exposure: (1) the constant of proportionality, and (2) the value of H
f
that will produce the desired density on the ﬁlm. It turns out that the constant
of proportionality can be obtained by a physical argument, so the only thing
remaining is to specify H
f
in terms of a ﬁlm.
Now black and white ﬁlms are characterized by the exposure, H
m
, required
to produce a density of 0.1 over background and fog. In fact if H
m
is measured
41
in cd/m
2
, then the ASA ﬁlm speed is deﬁned as S = 0.8/H
m
. Thus a fast ﬁlm,
requiring a small value of H
m
to produce a 0.1 density, will have a higher speed
than will a slower ﬁlm which needs a larger H
m
to produce a 0.1 density. It
follows then that one can use S to determine H
f
, if one is happy with H
f
being
deﬁned in terms of a density of 0.1 over background and fog.
This, however, is inconvenient, since measuring the luminance of shadows in
a scene can be diﬃcult. Therefore, H
f
is usually deﬁned in terms of a larger
density, near the middle of the range of interesting densities. For black and
white ﬁlm, the range of usable densities is taken to be 7 stops, hence one needs
to choose something like H
f
= rH
m
, where r is approximately 3.5. The upshot
is that L/H
f
becomes LS/K, where K is a called the calibration constant
and is used to encompass both r and the physical constant of proportionality
mentioned above. The details may be found in Jacobson(1988), pp 5052, and
pp 264268. If L is measured in cd/m
2
then K = 12.5 for the current ISO
standard.
In summary: one has a relationship between the camera controls, N and t
and the scene luminance L and ﬁlm speed S, expressed as
N
2
t
=
LS
K
,
which may seem more familiar in terms of the exposure value EV , deﬁned by
EV = log
2
(N
2
/t) = log
2
(LS/K). (52)
The ISO standard for color ﬁlm is deﬁned using the same calibration con
stant, but since color ﬁlm involves three ﬁlms, and since the range of color ﬁlm
is closer to 5 stops, the compromise is not always a happy one.
11.2 Light meters
The calibration problem for color ﬁlm is apparent for light meter calibration.
Newer meters, such as the newer Sekonic’s, use the current calibration constant
of 12.5. Older meters use other values. The Minolta Spotmeter F uses a cali
bration constant of 14. The Sekonic manuals state this explicity, but it is easy
to obtain for any meter that publishes a table converting EV into luminance.
The formula K = LS/2
EV
may be used with luminance L in cd/m
2
and S the
ASA ﬁlm speed. As mentioned in the previous section, calibration constants
determine the correspondence between densities and exposure. Increasing the
calibration constant, increases exposure and makes a thinner negative, which
some ﬁnd more appropriate for color ﬁlm.
Many users ﬁnd the calibration used by the Minolta Spotmeter F more ap
propriate for color ﬁlm, and on occasion have been know to claim that the
Sekonic’s are miscalibrated. There is clearly a diﬀerence between the two cal
ibrations over and above the calibration constant. Changing from K = 12.5
to K = 14, will decrease the EV by 0.16, however, I ﬁnd it necessary to de
crease the EV by about 0.3 to make the meters match. There also seems to be
diﬀerence in spectral response of the two meters.
42
11.3 18% gray cards
Since log
2
(100/18) ≈ 2.5, it follows that in any scene an 18% gray card will be
placed 2.5 stops down from the maximum brightness. If the scene has a range
of 5 stops, then the middle of the range will be 18% gray, and a lightmeter will
place the lowest tone one stop up from H
m
. For a 7 stop range, 18% gray will
be placed 4.5 stops up from H
m
and therefore an 18% gray will not be in the
middle of a 7 stop range.
11.4 Sunny 16 rule
The sunny 16 rule states that the correct exposure for bright sunlight should
be f/16 at 1/S seconds, where S is the ASA speed. The deﬁniton of exposure
value above, one has N
2
/t = SL/K, and if one takes t = 1/S, then one has
N
2
= L/K. For N = 16 this gives L=3200 cd/m
2
which is the proper value for
a midtone in bright sunlight. Interms of DIN, one has t = 10
(DIN−1)/10
.
11.5 Contrast control — the Zone system
The standard assumes an exposure range of twice 3.6 EV, or about 7 EV. If the
luminances in a scene span this or a smaller range, then all will fall on a usable
portion of the characteristic curve. For longer ranges, the lower portions of the
range will not be captured by the ﬁlm and both extremes may be lost when
printing: the diﬃculty is less serious for B&W ﬁlm with its greater latitude.
For a range smaller than 7 EV, one may adjust the standard by decreasing
the EV. For larger ranges this will not help. For B&W ﬁlm, one may change
the characteristic curve by changing processing times, and this may be used to
adjust the range to the range of the paper.
The Zone System is a methodology for controling the contrast of B&W
negatives. Adams (1981) assigns 10 zones as shown in Table (4). In practice,
the seven zones II through VIII are the only ones of interest. The object is to
expose and process ﬁlm so that the luminance range of the subject is capatured
within this seven zone range.
0 Pure paper black
I Black; virtually indistinguishable from 0
II Near black; texture but no detail
III Very dark gray
IV Dark gray
V Middle gray
VI Light gray
VII Very light gray
VIII Near white; texture but no detail
IX Pure paper white
Table 4: Zone deﬁnitions
43
Davis (1999) describes a systemaic methodology for adjusting the contrast
for B&W. He uses the subject luminance range, which he designates by SBR,
as a basis. If ∆E is the range of EV’s in the scene to be mapped into a range
∆Z of zones, then he deﬁnes SBR = 7/r, where r = ∆Z/∆E. The assumption
being that 7 is a normal range. He multiplies the average gradient, G by r to
obtain the average gradient required of the characteristic curve in order to place
the zones in an acceptable density range for the negative. Experiental eﬀort is
required to ﬁnd suitable processing parameters to achieve G.
The appropriate exposure is obtained from
EV = (loEV ) −(loZone −1.5)/r + 3.6 + log
2
(r), (53)
where loEV is the EV to be placed on the zone loZone. The argument leading
to equation (53) is that since the light meter calibration adjusts EV’s by 3.6 to
produce mid gray tones on a seven stop scale, one can remove this calibration
to obtain a base level corresponding to an exposure which will produce a just
noticeable tone: presumably this corresponds to zone I+1/2. Since each zone
corresponds to 1/r EV units, one must then add (loZone −1.5)/r to obtain the
EV that will produce a tone corresponding to lo Zone.
In addition, one must allow for the eﬀect of speed, which changes according
to the SBR. If ASA
7
is the ASA speed when SBR = 7, then the ASA speed
should be adjusted to r ×ASA
7
.
Davis apparently increments from zone I1/2, and uses 4 instead of 3.6 to
obtain equaion (54),
EV = (loEV ) −(loZone −1/2)/r + 4. (54)
The diﬀerence between equation (53) and (54) is log
2
(r) − 0.4 − 1/r, which
is negative for r < 1.9, resulting in overexposure for developing times that
compress the range. The lighter tones will appear to be blocked using Davis’s
formula.
11.6 Characteristic curves
11.6.1 Graduating function
The following function may be used to describe a characteristic curve:
D(x) = F +D
M
sin
π
2
x −I
R
2
, (55)
where D is density, D
M
is the maximum density, x is exposure (log H), I is the
intercept, R is the exposure range, and F is base plus fog.
The gradient at x is
D
(x) = D
m
π
R
sin
pi
2
x −I
R
cos
pi
2
x −I
R
,
and the average gradient over [I, I +R] is
44
¯
G =
D
M
R
sin
π
2
x −I
R
2
I +R
I
=
D
M
R
, (56)
which diﬀers slightly from that used by Davis (1999), where only the center
range of densities are used.
11.6.2 Data
Test 1 Test 3 Test5
F 0.077 0.069 0.056
D
M
0.796 1.792 2.531
I 2.934 3.176 3.220
R 3.805 4.131 3.506
Speed 2.06 2.55 2.77
Davis’s Speed 2.2 2.6 2.8
¯
G 0.21 0.43 0.72
¯
G
c
0.27 0.57 0.86
Davis’s
¯
G
c
0.31 0.56 0.88
SBR 9.8 5.8 3.9
Davis’s SBR 10.7 5.9 3.8
Table 5: Fitted parameters
The data in table (6) was copied from page 47 of Davis (1999), and ﬁtted
with equation (55). Figure (23) shows a plot of the ﬁtted values. The ﬁt is
essentially a perfect ﬁt, suggesting that the graduating function captures the
essence of the underlying mechanism. Similar plots with other ﬁlms show the
same good agreement.
Table (5) gives the ﬁtted parameters plus other derived estimates. Film
speed is obtained as the log H value corresponding to a density of 0.1+F. The
¯
G is calculated from equation (56), and
¯
G
c
is the average gradient obtained
from the center range of densities from 0.1+F to DR+0.1+F using equation
(55), where DR is the useable density range which is approximately equal to
the paper’s SI: here the DR is taken as 1.0. The subject luminance range, SBR,
deﬁned by Davis is DR/(
¯
G
c
log(2)).
It is important to note that the maximum density predicted by the grad
uating function is M, which in the case of Test 1 is less than DR+0.1+F. In
this case
¯
G
c
is calculated using the density range from 0.1+F to M. Since the
graduating function maximizes at this point, it suggests that the development
process will not achieve higher densities, and than greater exposures will pro
duce a decreasing density solarizaiton. The SBR is therefore taken as M/
¯
G
c
45
Figure 23: Fitted values
12 Perspective and normal lenses
Without the ability to assess distance using binocular vision, the visual system
must rely on other clues: the principal one being angular size. The assumption
that the angular size of an object is proportional to distance is not a bad one.
The ratio of object height, h, to distance, s, is h/s = (h/f)[m/(1 + m)] which
is approximately m(h/f) for small m, where f is focal length, and m is mag
niﬁcation. Thus twice m(h/f) means either that the object is twice the size
m(2h/f) or twice as close 2m(h/f), since 2/s ≈ 2m/f. For small angles, one
has arctanh/s ≈ h/s and thus the proportionality also applies to angles.
The human eye is not a camera, but it does obey the laws of optics, and it
is interesting to discuss it as if it were a camera. Such a camera would have a
focal length of about 1
and the image projected on the retina would also be
1
. It follows that the angle of view, AOV, is 2 arctan(1/2) = 53
◦
.
Let D
n
be the nearest distance of clear vision, which seems to be in the
neighborhood of 10
for normal eyes. The magniﬁcation is thus m = 1/D
n
, and
an object ﬁlling the retina will be of size D
n
. Since m is small, the proportion
ality holds for the entire practical range of the visual system: for all objects
from the nearest focus to the most distant. Cameras are not limited to small
m, and nearby objects which violate proportionality, can be brought into focus,
enabling pictures in situations to which the eye is unaccustomed.
46
log H Test1 D Fit 1 Test 3 D Fit 3 Test5 D Fit 3
3.05 0.08 0.08 0.09 0.07 .10 0.07
2.9 0.08 0.08 0.09 0.09 0.11 0.11
2.75 0.08 0.09 0.11 0.12 0.16 0.17
2.6 0.09 0.09 0.14 0.15 0.23 0.25
2.45 0.10 0.11 0.19 0.20 0.32 0.34
2.3 0.12 0.13 0.25 0.26 0.44 0.46
2.15 0.15 0.16 0.33 0.33 0.58 0.59
2 0.19 0.19 0.41 0.40 0.75 0.74
1.85 0.23 0.23 0.50 0.49 0.92 0.90
1.7 0.27 0.27 0.59 0.58 1.08 1.06
1.55 0.32 0.31 0.68 0.67 1.25 1.23
1.4 0.36 0.36 0.77 0.77 1.41 1.40
1.25 0.41 0.40 0.87 0.87 1.58 1.57
1.1 0.45 0.45 0.97 0.97 1.73 1.73
0.95 0.50 0.50 1.07 1.07 1.88 1.89
0.8 0.55 0.55 1.17 1.18 2.01 2.03
0.65 0.59 0.60 1.27 1.27 3.15 2.17
0.5 0.64 0.64 1.36 1.37 2.27 2.29
0.35 0.68 0.69 1.45 1.46 2.38 2.39
0.2 0.73 0.73 1.54 1.54 2.48 2.47
0.05 0.77 0.76 1.62 1.61 2.55 2.53
Table 6: Observations and ﬁtted values
To appear normal, a picture held at D
n
should produce the same angles
for objects as does the image that forms on the retina. An image occupying a
proportion, α, of the picture will have an angle of 53α
◦
. If a camera lens has
this AOV, then an object occupying the proportion α will also have angle 53α
◦
,
and thus appear normal on a print. It follows then that normal lenses are those
whose focal lengths are approximately equal to the diagonal of the ﬁlm size.
A lens with a smaller than normal focal length will have an AOV greater than
53
◦
, making an object occupy a smaller angle and thus appearing farther away.
A larger focal length does the opposite, making distant objects appear closer:
the consequence of this is that two objects which to the naked eye appear X units
apart, will appear to be to be X53/A
L
units apart, where A
L
is the AOV for the
longer lens. For shorter lenses with A
S
, the AOV, the situation is reversed, so the
objects appear to be XA
S
/53 units apart: the separation appears greater than
the naked eye expects. In addition, short lenses are more frequently used for
nearby objects for which proportionality does not hold, producing even greater
exaggeration. It should be noted, of course, that such nearby objects, even when
they cannot be brought into focus, still appear reasonable to the eye because
binocular vision provides triangulation information to the visual system.
47
13 Digital
13.1 Scanning and resolution
The aim of scanning is to extract that amount information from the source
material required by the output material. How much is this? Obviously it
varies with the purpose, but for photographs, it need not be more than the
human eye can use. The minimum distance of distinct vision is about 25 cm.
Which means that most people cannot focus on closer objects. In bright light
the human eye at this distance can just resolve 5 l/mm which sets a limit on
the print resolution required. What is the corresponding resolution on the ﬁlm?
There are two parts to the problem: (1)Scan to digital, (2) Digital to print.
(1) A careful examination of scans a USAF test target, suggests about 3 dots
seems to be the minimum number required to separate lines. A line on these
charts is the extent of a black and white bar, which is the distance between the
frontiers of one line and its successor. With two dots, the line may be missed
because both dots fall in the black or in the white, while with 3 there will be
at least one on each color. It follows that for a scanning density of R dpi, the
captured ﬁlm resolution r
f
is about R/3 l/in, or about R/75 l/mm.
(2) The density, R, must be determined from the print requirements. Now
5 l/mm, the resolution of the human eye, corresponds to 127 l/in, but printers
need 2 input dots to show a USAF test target line, this means that about 250
dpi must be sent to the printer. As conﬁrmation, tests using photographs of
a USAF test target with the Epson Photo EX shows that this printer requires
a value between 200 and 240 dpi. The 240 dpi is excellent, and the 200 dpi is
almost as good, but a slight degradation can be detected. Large commercial
printers often require 300 dpi. For convenience, let us choose 225 dpi as the
required dot density for a printer.
The total number of dots required by a printer is T = 225 × d
p
= R × d
f
,
where d
p
and d
f
are corresponding dimensions of the print and the ﬁlm. From
this, one has R = T/d
f
, and thus the captured r
f
becomes r
f
= R/75 =
(T/d
f
)/75 = (225d
p
/d
f
)/75 = 3M in l/mm, where M = d
p
/d
f
. If one chooses
2 dots instead of 3 in paragraph (1), then r
f
= 4M or 5M. If one sends 300 dpi
to the printer, as one would to with a LightJet 5000, then r
f
= 4M.
Taking r
f
= 3M, the captured r
f
is shown in Table 7 for several ﬁlm and
print sizes. Appropriate scan rates for each ﬁlm size and the maximum onﬁlm
captured r
f
is also shown.
4x5 8x10 16x20 R R/75
35 12 24 51* 3000 40
6x7 6 12 24 2000 27
4x5 3 6 12 1000 13
Table 7: Captured ﬁlm resolutions in l/mm given print size
* Not possible with available scanning rate
48
One sees, for example, that the minimum captured detail in an 8x10 digital
print from a 4x5 transparency is about 6 l/mm. This is less than the minimum
resolution for a conventional print of about 10 l/mm. The diﬀerence is traceable
to the fact that 3 dots are needed by the scanner for a digital print. The loss
can often be compensated by sharpening the edges with unsharp masking.
It is useful to note that four times the 4x5 r
f
is equal to the 35 mm r
f
.
Therefore given a ﬁxed print size, the resolutions for all formats represent the
same degree of detail in the subject: e.g. an object just resolved in an 8x10
print from a 35 mm slide will be just resolved in an 8x5 print from a 4x5. This
is not true for a 16x20 print, where there is a diﬀerence between 35 mm and 4x5
because the 35 mm r
f
is limited to 40 l/mm (= 3000/75) instead of 48 l/mm.
As a caution, there is always much more resolution in a ﬁlm than may be
captured in a scan. Figure 17 on page 33, shows a stack of dollars which were
photographed and scanned at 3000 dpi — it is printed here only to show the test
setup, since the resolution is low in this printing. The photo used was 1322 in
my ﬁles. The resolution of the ﬁlm is near 60 or so l/mm, since the major ﬁne
lines at 45 l/mm are clear with a 20 power loupe, but the lines in the leaves at
about 72 l/mm are not distinguishable
12
.
In the scanned image, the major ﬁne lines can be seen against a black back
ground, but they cannot be separated when near each other as in the wreaths,
nor can the lines in the leaves be separated. I conclude that one needs more than
two dots to make a line clearly distinguishable. If one requires 3 dots instead
of 2, then the resolution from a 3000 dpi scan is 3000/150 = 20 l/mm, which
seems reasonably consistent with these observations. Scanned photographs of
the USAF chart conﬁrm that 3 dots are needed at the highest detectable fre
quency.
13.2 Digital cameras
The arrays in digital cameras are much smaller than ﬁlm sizes. The commonly
used ones are shown in Table 8. The dimensions have been rounded to the
nearest millimeter. The names 1/3 inch, etc. are a holdover from previous
technology and do not represent the actual size of the arrays.
height width pixels
1/3 inch 4 5 480x640
1/2 inch 5 6 1000x1300
2/3 inch 6 9 1400x1700
large 18 28 2000x3000
Table 8: Approximate digital camera array sizes in mm
12
Comparing the line width with that of an Edmund chart shows the major lines are about
5 l/mm, and the ones on the leaves perhaps 8 l/mm. A 150 mm lens was used at a distance
of 1500 mm, giving a multiplying factor of 9, so 9 ×5 = 45.
49
3x5 5x7 8x10 R R/75
1/3 inch 57* 95* 152* 3200 43
1/2 inch 45 76* 122* 5000 67
2/3 inch 38 63 102* 6000 80
large 13 21 34 2820 38
Table 9: Captured digital array resolutions in l/mm given print size
* Not possible with available array dimensions
A table of captured resolutions for these cameras is shown in Table 9. The
R in this table is the number of pixels per inch for the shorter dimension of the
array. It may be seen that only the large array is adequate for 8x10 prints. A
limited test seems to conﬁrm the accuracty of this table. Tests on the Olympus
C2500L with a 2/3 inch array produced 60 to 68 l/mm depending on the focal
length used. This agrees with the captured resolution in Table 9 of 63 l/mm.
14 Miscellany
14.1 Fractional stops
The amount of light admitted through the aperture is of great importance, as
early photographers quickly learned. It may be controlled by the duration of the
exposure and by the lens aperture. About 1880 according to (Kingslake, 1989)
it became customary to designate a lens by its fnumber. Iris shutters were also
introduced about 1880 (ibid), and markings on the lenses were likely to have
been in terms of fnumbers. Sequences of stops and fnumbers were proposed
at the turn of the century, but the present standard sequence of fnumbers {1,
1.4, 2, 2.8, 4, . . .} is fairly recent.
There is not a lot that can be said in justiﬁcation of this standard sequence
for a practical photographer. It requires an act of memory to use, and it in
creases as the light admitted decreases, which is contrary to what one wants.
One way to think of it is as a measure of the luminance of the object, in which
case a small opening and large fnumber represents a bright object, but the odd
sequence is still something to remember.
A rational scale of values corresponding to aperture diameters
13
is the expo
sure value, EV , scale. From equation (52), one can see the relationship between
EV and the fnumber, N by setting the time t to unity: it is N
2
= 2
EV
, or
EV = 2log
2
(N).
The EV value 2.5 corresponds to a point between f/2 and f/2.8. It is this
13
The blades of iris diaphragms are curved to produce an aperture scale with equidistant
spacings, (Ray,1994). The movement may not be completely uniform for intermediate stops,
since the opening is a polygon, not a circle. In general fractional stops on large format lenses
should not be relied on to less than 1/3 stop and probably not to 1/2 stop.
50
“fractional” value that appears on digital light meters
14
: e.g. a meter reading
might be 1.45 with the “5” in smaller type than the “1.4”. Useful notations for
the printed page are 1.4 5, and 1.4 +1/2. Cameras with automatic diaphragms
can be expected to adjust their apertures to this scale, at least to a 0.1 step
(Ray, 1994).
The amount of light actually admitted through the diaphragm is not, how
ever, %50 more at EV = 2.5 than at EV = 2.0. From equation (52), one has
2
EV
= LS/K. Thus 2
EV +0.5
= (LS/K)2
0.5
= (LS/K)
√
2 ≈ (LS/K)1.41; thus
the increase is %41 not %50.
When magniﬁcation, m, is not zero, one may desire the number of stops
associated with the bellows factor. If V is the fnumber when focused at inﬁnity,
then U = V/(1+m) is the fnumber allowing for the bellows extension. Suppose
EV
1
is the exposure value corresponding to V , that is 2
EV
1
= V
2
, then U
2
=
2
EV1
/(1 + m) = 2
EV1−log2(1+m)
. Thus one opens up log
2
(1 + m) stops
15
from
V to obtain the correct fnumber U.
14.2 Stopping Motion
Let c be the diameter of the circle of confusion, and s the corresponding subject
size at distance u for a lens with focal length f, then
c
s
=
f
u −f
,
and if an object moves at a rate r then s = r/T, where 1/T is the shutter time
in seconds, one ﬁnds that the shutter speed necessary to stop this motion is
T =
rf
c(u −f)
= m
r
c
,
where m is magniﬁcation and r is in millimeters per second.
If r is in mph, u in meters, and f and c in millimeters, then
T =
447.04rf
c(1000u −f)
.
It may also be useful to cite the distance the image of a point moves during
1/T seconds. This is given by
mr
cT
,
where the distance is in units of c.
14
See the Minolta Spotmeter F instruction manual, or the tables in the Sekonic Zoom Master
L508, or L608.
15
Remember that a decrease in EV corresponds to a decrease in fnumber, and an increase
in aperture size.
51
14.3 Bellows Factor
The inverse square law applies to bellows movements: the intensity of the light
falling on a surface is inversely proportional to the square of its distance from
the lens. Thus a bellows with extension equal to the focal length of the lens
delivers
1
4
the light, or exposure, as one with zero extension.
If e is the extension of the bellows, then the bellows factor is deﬁned as
BF =
f +e
f
2
= (1 +m)
2
,
where m is magniﬁcation.
The time value (1/sec) is multiplied by BF to adjust for the decrease in
light, or the fnumber can be divided by
√
BF.
An easy procedure is to divide f + e and f by something that brings them
into the conventional fnumber range and then to count the stop diﬀerences:
thus, e = 250 with f = 200 produces the 45 and 20 fnumbers, which is a bit
more than 2 stops. This may be compared with the actual value of 2.3 stops
given by log
2
(BF).
14.4 Stops as distances
The inverse square law gives the ratio (d
2
/d
1
)
2
for light intensity at two dis
tances d
1
and d
2
, and the number of stops is the base 2 log of this ratio; hence
2log
2
(d
2
/d
1
) is the change in stops.
14.5 AOV
The angle of view, A, is the angle subtended at the center of the lens by either
the object ﬁeld or the image extent. Simple trigonometry shows:
A = 2 tan
−1
d
2f(1 +m)
,
where d is the image extent, m the magniﬁcation, and f the focal length.
It is useful to note that this may be rearranged as,
1 +m =
d
2f tan(A/2)
,
which shows that if f is scaled by the same scaling as d when format is changed,
then magniﬁcation is unaﬀected for a constant AOV. The camera must be moved
to accommodate this, since
v = f(1 +m) =
d
2 tan(A/2)
,
changes with d.
52
14.6 Film ﬂatness and vibration eﬀects
The eﬀect of a lack of ﬁlm ﬂatness or vibration may be assessed in terms of
changes in the depth of focus. Equation (34) may be used with c = 1/L:
L =
2N(1 +m)
δ
,
where δ is the depth of focus along the rail, and L is lines per millimeter. The
diﬀerencial of δ with respect to L becomes
dδ = −
δ
L
dL = −δc dL,
which is the amount of ﬁlm movement which degrades the image by one line
per millimeter — the smaller the dδ the more sensitive. One should probably
use half of this, since the ﬁlm backing is presumably ﬁxed.
This suggests that ﬁlm movement disturbs the image at the exact focus more
than at the DOF limits.
14.7 Hand holding a camera
Taking 1/f as the exposure time is a common rule for hand holding. It has
some justiﬁcation: First assume that halving the time of exposure will halve
the size of the blur. Fix the magniﬁcation so that the subject will always image
the same size on the ﬁlm. Suppose that vibrations from a lens of focal length
f, can be damped by an exposure time of 1/f . This means that a point in the
subject will image as a circle of confusion, c. It follows that a 2f lens puts the
image twice as far from the lens, and with the same amount of vibration, the
image will be of size 2c.Hence one should cut the vibrations in half by using
1/(2f) as the time of exposure to obtain an image of size c.
A test must be run to justify 1/f in the ﬁrst place. It is likely that a
parameter will be needed to reﬂect individual variation.
14.8 Finding the 1:1 focus point
Since u + v =
u
2
u−f
which has a minimum at u = 2f, one sees that the 1:1
focus point is attained when u + v is a minimum. In practice one could use a
measurement from a mark on the rear standard to a mark on something that
moves with the subject and vary the subject distance until the minimum infocus
distance is found.
14.9 Miscellaneous equations
Magniﬁcation as a function of nearDOF.
m =
Nc(f −∆)
f∆
+
Nc(f −∆)
f∆
2
+
Nc
∆
, (57)
53
where ∆ = u −Z
n
is nearDOF.
Focal length as a function of DOF.
f =
Nc
m
DOF
DOF −2NcM
, (58)
where M = (m + 1)/m
2
.
54
BIBLIOGRAPHY
1. Adams, Ansel. (1981) The Negative. Little Brown, Boston.
2. Davis, Phil. (1999). Beyond the Zone System. Focal Press, Oxford
3. Hansma, Paul K. (1996) View camera focusing in practice. Photo Tech
niques. 172, 5457.
4. Jackobson, R.E., Ray, S.F., and Attridge, G.G. (1988). The manual of
photography. Focal Press, Oxford.
5. Kingslake, R. (1979) Lens design fundamentals. Academic Press, Boston.
6. Kingslake, R. (1989) A history of the photographic lens. Academic Press,
Boston.
7. Merklinger, Harold, M. (1993) Focusing the view camera. Ottawa, Canada.
8. Mouroulis, P. and Macdonald, J. (1997) Geometrical optics and optical
design, Oxford, NY.
9. Ray, S.F. (1994) Applied photographic optics, Focal Press, Boston.
10. Sears, F.W. (1958) Optics, Addison Wesley, Reading MA.
11. Stroebel, Leslie; Compton, John; Current, Ira; and Zakia, Richard. (1986)
Photographic materials and processes. Focal Press, Boston.
55
Contents
1 Desargues’s Theorem 2 The Gaussian Lens Equation 3 Thick lenses 4 Pivot Points 5 Determining the lens tilt 5.1 Using distances and angles . 5.2 Using back focus . . . . . . 5.3 Wheeler’s rules . . . . . . . 5.4 Lens Movement . . . . . . . 5.5 Back Tilts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 6 8 9 10 10 12 13 14 14 15 15 17 17 18 19 19 20 21 21 21 22 23 24 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 26 27 29 31 32 33 35 35 35 37 37
6 Depth of ﬁeld for parallel planes 6.1 Near DOF limit . . . . . . . . . . 6.2 Far DOF limit . . . . . . . . . . 6.3 DOF . . . . . . . . . . . . . . . . 6.4 Circles of confusion . . . . . . . . 6.5 DOF and format . . . . . . . . . 6.6 The DOF equation . . . . . . . . 6.7 Hyperfocal distance . . . . . . . 6.8 Approximations . . . . . . . . . . 6.9 Focus given near and far limits . 6.9.1 Object distances . . . . . 6.9.2 Image distances . . . . . . 7 Depth of ﬁeld, depth of focus 8 Fuzzy images 9 Eﬀects of diﬀraction on DOF 9.1 Theory . . . . . . . . . . . . 9.2 Data . . . . . . . . . . . . . 9.3 Resolution . . . . . . . . . . 9.4 Format considerations . . . 9.5 Minimum aperture . . . . . 9.6 Theoretical curves . . . . .
10 Depth of ﬁeld for a tilted lens 10.1 Near and far DOF equations . . . . . . . 10.2 Near and far DOF equations in terms of ρ 10.3 Focus given near and far limits . . . . . . 10.3.1 Object distances . . . . . . . . . .
2
10.3.2 Image distances . . . . . . . . . . . . . . . . . . . . . . . . 10.4 The graph of the DOF limits . . . . . . . . . . . . . . . . . . . . 10.5 Variation of maximum diameter of the ellipse of confusion with ρ 11 Image control 11.1 Exposure . . . . . . . . . . . . . . 11.2 Light meters . . . . . . . . . . . . 11.3 18% gray cards . . . . . . . . . . . 11.4 Sunny 16 rule . . . . . . . . . . . . 11.5 Contrast control — the Zone system 11.6 Characteristic curves . . . . . . . . 11.6.1 Graduating function . . . . 11.6.2 Data . . . . . . . . . . . . . 12 Perspective and normal lenses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
38 38 39 41 41 42 43 43 43 44 44 45 46
13 Digital 48 13.1 Scanning and resolution . . . . . . . . . . . . . . . . . . . . . . . 48 13.2 Digital cameras . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 14 Miscellany 14.1 Fractional stops . . . . . . . . . . . 14.2 Stopping Motion . . . . . . . . . . 14.3 Bellows Factor . . . . . . . . . . . 14.4 Stops as distances . . . . . . . . . 14.5 AOV . . . . . . . . . . . . . . . . . 14.6 Film ﬂatness and vibration eﬀects 14.7 Hand holding a camera . . . . . . . 14.8 Finding the 1:1 focus point . . . . 14.9 Miscellaneous equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 50 51 52 52 52 53 53 53 53
3
Preface
The following is a collection of notes and derivations that I made while reading about view cameras. Everything is geometrical (trigonometrical), and thus may not apply in various ways to actual lenses; although it should be pointed out that while actual lens’s may not attain these ideals, it is these for which lens designers strive.
1
Desargues’s Theorem
There are three fundamental planes: the ﬁlm plane, FP, the lens plane, LP, and the subject plane, SP. An object in SP will be in focus on FP only when the three planes bear a particular relationship. This relationship, interestingly enough, may be derived from a theorem of Desargues (15931662) in projective geometry. Projective geometry is the name for that branch of mathematics that arose to explain the relationships of three dimensional objects, projected onto a two dimensional canvas; these relationships were observed by artists in the Renaissance (about 1400) when they began to develop an understanding of perspective. To draw a rug on a ﬂoor, for example, whose sides appear proper, one must choose a vanishing point, VP, on the horizon at which lines through the sides of the rug converge. The interesting thing is that if one draws a object, say box in three quarter view, then lines through its sides must converge to VP’s on the horizon and all such VP’s must lie on the same line. This convergence to points on a line is the essence of Desargues’s theorem. Theorem (Desargues): If in a plane two triangles ABC and abc are situated so that the straight lines joining corresponding vertices are concurrent in a point O, then the corresponding sides, if extended, will intersect in three collinear points. This is illustrated in space by ﬁgure 1. The theorem is rather diﬃcult to prove for triangles lying in the same plane, but it is quite easy to prove if they lie in diﬀerent planes. In ﬁgure 1, aA, bB, cC are concurrent at O according to the hypothesis. Now ab lies in the same plane as AB, so that these two lines intersect at some point P2; likewise ac and AC intersect at P3, and cb and CB intersect at P1. Since P1, P2, and P3 are on extensions of the sides of abc and ABC, they lie in the same plane with each of these two triangles, and must consequently lie on the line of intersection of these two planes. Hence P1, P2, and P3 are collinear which was to be proved.
4
O
A B c a b
C
P3
P2 P1
Figure 1: Desargues’s conﬁguration in space It should be noted that the theorem holds, and the above proof follows through if one places the concurrent point O at inﬁnity, as illustrated by ﬁgure 2, which brings us to Photography, since the three lines SP, LP, and FP in ﬁgure 2 represent the three photographic planes mentioned above. P1 may also be at inﬁnity, in which case the three planes are parallel. The point b in ﬁgure 2 is the lens center, and the lines P2a, and P3c are rays through the lens from SP to FP. The triangles P2Aa and P3Cc are those usually used to graphically depict image formation — see Stroebel, et.al. (1986). The lines aA and bB1 and cC are drawn to be orthogonal to LP. Hence the extensions of aA, cC and bB are coincident at a point at inﬁnity, satisfying the condition of Desargues’s theorem, and thus the points P1, P2, and P3 are collinear. Since such a diagram may be drawn for any three lines which meet at a point, one has Scheimpﬂug’s rule. Scheimpﬂug’s rule: The subject plane is rendered sharp when the subject plane, the lens plane, and the focus plane all meet in a line.
1 B may be placed anywhere on the line orthogonal to LP as far as Desargues’s theorem is concerned, but for Photography, it should be placed one focal length away from b, making it the object focal point.
5
f. equation (1) may be obtained as follows: 1 1 + u v = = = 1 1 + . We will now show that Desargues’s theorem also implies the Gaussian lens equation: 1 1 1 + = . The lines Aa. is deﬁned. and the dashed line parallel to SP ending at point c is a line from inﬁnity: since this line meets SP at inﬁnity Desargues’s theorem is satisﬁed. the three planes satisfy the Scheimpﬂug rule. P b sin α P a sin β 1 1 + . The point L is the lens center. The length of Cc. Bb and Cc are orthogonal to LP. f . u v f Using ﬁgure 3. as the focal length of the lens. f f P b cL P aPc 1 f cL Pc + Pb Pa .SP P3 LP A b B P2 FP a c C O P1 Figure 2: The three planes 2 The Gaussian Lens Equation Figure 3 shows the three planes from ﬁgure 2 with some additional lines and points. (1) and since by similar triangles cL/P b = ca/P a. As has been noted. one has 1 1 1 + = u v f ca Pc + Pa Pa 6 = 1 . as is customary.
(2) SP LP v y FP a A L h α u B f ω c b C α β P Figure 3: The lens equation 7 .One can express this in terms of distances along the line ab as 1 1 cosω + = h y f where u = hcosω and v = ycosω.
Because the lens is thick. Camera lenses are not thin. Fortunately. In air. but the diﬀerences are usually so small as to have no eﬀect on the acceptable precision of the calculations. and is an approximation to a thin lens. 8 . where α is the 2 The nodal space or hiatus is the thickness of the lens as measured by the diﬀerence between the two nodal points. u f u−f (5) u v = = (6) (7) and of course the Newtonian form: f 2 = (u − f )(v − f ). (8) These equations hold for h and y when f is replaced by f /cosω as per equation(2). and the focal length.the thickness can be ignored. the consequences are negligible. v−f uf . is constant for all values of ω. use will be make of the fact that magniﬁcation. see (Kingslake 1978) or (Mouroulis and Macdonald 1997). This follows algebraically and by inspection of the similar triangles in ﬁgure (3). Light from a lens may be described geometrically as if it radiated from a single surface. The geometry is attributed to Gauss. then u becomes u = u − δ/ cos α ≈ u − δ. m f (1 + m). the diﬀerence between the principal planes is equal to the nodal space. f (1 + v v−f f = = . distance v. If δ is the nodal space2 . which is an abstraction describing a lens such that.There are several derived forms of equation (1) that are frequently of use. Later. The front principal plane is cognate to the rear principal plane in that they are images of one another with unit magniﬁcation. u v = = vf . This plane acts like a plane though a thin lens. u−f (3) (4) And if one deﬁnes magniﬁcation by m= then 1 ). which is essentially a plane for paraxial rays called the rear principal plane. m. within acceptable precision. variables on the object side are not quite what is given by the thin lens geometry. 3 Thick lenses The above is geometrical. and other variables on the lens side act as in the geometric theory above.
in fact. J and K are the lengths of the rays from the lens center. P must be f units from LP. By deﬁnition.angle of a ray with respect to the lens axis. and FP is moved parallel to itself. The points P and Q are pivot points. It only becomes important when δ is substantial with respect to u — it should be noted that δ can be negative. A similar statement may be made about the reciprocal pivot point Q with respect to movements of SP parallel to itself. as at ﬁrst might be thought. Later it will be seen that the near and far depth of ﬁeld limits are essentially planes corresponding to movements of FP parallel to itself. hence for u = 10 meters. and since there is only one such point on the segment of IP below L. to be satisﬁed. to P and Q respectively. IQ IP SP LP θ φ FP L f f φ θ K f J γ f Q P V Figure 4: Pivot points Suppose LP is ﬁxed. nor does it pivot about V. this diﬀerence hardly matters. equation (1). as will now be discussed. P is the point where IP meets SP. and δ = 2 centimeters. It. the correction is u = 9.96 meters. 4 Pivot Points Figure 4 shows the three planes and two lines from inﬁnity: the line IQ meeting FP at Q and the line IP meeting SP at P. then SP must also move in order for the lens equation. L. Even for close up photography. The distances from P and Q to LP must of course be f. it follows that SP must pivot on P as FP is moved parallel to itself. SP does not move parallel to itself. turns about the pivot point P. hence they too must 9 .
of length Z. This can sometimes be done. since one does not need to know φ very precisely — within a degree is usually satisfactory. Consider ﬁgure 5 where two rays of length d and D have been drawn from the lens center on LP to SP. A more rigorous procedure is to measure the distance and angle of two points in SP. Merklinger(1993).1 Determining the lens tilt Using distances and angles There are some authors. who suggest that φ. the lens tilt angle. The tilt angle of the lens. with respect to FP is easily seen to be given by arcsin(f /J). The longer ray is B degrees above the lens axis. SP Y X x D LP FP Z y d Β β K J γ P V Figure 5: J from two rays 10 . Similarly the tilt with respect to SP is θ = arcsin(f /K). from the lens center and the shorter one is β degrees below the horizontal. 5 5. the angle of SP with respect to FP. and then calculate J.pivot on P. and γ. φ. should be determined by guessing J. is φ + θ.
J D sin B. D sin B + J Although not needed here. D cos B − Z JD cos B . then one may eliminate Z by using two rays and equating the values of equation (10) for each: Jd cos β d sin β + J dD(sin B cos β − cos B sin β) dD sin(B − β) 1 J = = = = JD cos B . D cos B − Z. δ sin α + J J Z = = (9) (10) for an arbitrary ray of length δ and angle α. D sin B + J J(d cos β − D cos B). A similar calculation may be made for the ray below the lens axis. it is useful to note that D = Z cos(γ)/ cos(B +γ). J(d cos β − D cos B). but one must adopt a convention about signs. dD sin(B − β) One may then ﬁnd φ. from sin φ = f (D cos B − d cos β) . ZD sin B . then one may ﬁnd Z by equating the values from equation 9 for two rays. obtaining: 11 . then from equation (9). hence we will assume that angles below the lens axis are negative. SP makes with respect FP will be needed shortly. d cos β − D cos B . γ. one has: X Y Y X J Z = = = = = Z . If Z is known. tan γ = (δ cos α − Z)/δ sin α. using φ = arcsin(f /J). If one cannot. It seems better in the present case to reverse this convention. dD sin(B − β) (11) The angle. This leads to a common formula for both types of rays: Zδ sin α . or chooses not to measure Z. If Z is not known.For the ray above the lens axis. It is given by γ = arctan(Z/J). then one may use equation (9) to ﬁnd J and hence φ = arcsin(f /J). If one can measure Z. There is a standard convention which chooses B negative and β positive. δ cos α − Z Jδ cos α .
is used in subsection 5. on SP are shown in focus on TFP at a and c respectively. Thus the distance from A to LP is conjugate to the distance from LP to a.Z= and then.2 Using back focus Figure 6 shows both a tilted lens plane (TLP). D sin B − d sin β D cos B − d cos β . Thus one may locate a and c by moving FP. It is the conjugate of Z for an untilted lens: K= f . Two points. Z −f 5. and these distances are the same both for a tilted ﬁlm plane and for a vertical ﬁlm plane.2. d sin β − D sin B (12) tan γ = The distance K shown in ﬁgure 5. and a tilted ﬁlm plane (TFP). The angle α is given as the arctangent of (b − a)/(c − b). and it is easily seen that φ may be found from 12 . SP A u C TLP LP FP TFP φ c K J a α b γ α f Figure 6: Back movements to obtain tilt Note that the conjugates of SP points are on both TFP and on vertical planes through the conjugates. A and C. dD sin(B − β) .
and must be focused by moving the lens. hence one may often take φ = α. one may ﬁnd γ from tan(γ) = which leads to: sin(φ) = 1 tan(α) Kf 1 f 1 tan(α). 5. In other words sin(φ) is half the harmonic mean of tan(α) and tan(γ). = tan(α) = K −f J K −f m 1 + 1 tan(γ) . Let T be the distance between the top of the front standard and the tilt axis. the two images are about 60 13 . The rail distance between the two points of focus on the camera is 12 mm. On the groundglass. one can calculate the tilt in terms of the distance ∆ that the top of the front standard moves. K f 1 = . then Wheeler’s tilt rule is ∆=T (b − a) (u − f ) . K cos(α) (1 + m ) cos(α) where K = uf /(u − f ) = f (1 + m ) with m the magniﬁcation for a ray through the untilted lens center. (c − b) u A more practical expression for use in the ﬁeld is Wheeler’s rule of 60. given by φ ≈ 60 where (u − f )/u is close to unity except for closeup work. In general Q is not greatly diﬀerent from unity.3 Wheeler’s rules (b − a) (u − f ) . Finally. Since some cameras do not have angle scales. Some cameras do not have movable backs.sin(φ) = where Q= f tan α = Q sin(α). See ﬁgure (20). (c − b) u where (u − f )/u is close to unity except for closeup work. The distance from the lens center to FP is of course uf /(u cos(φ) − f ) because ω in equation (2) is φ. EXAMPLE: A plane containing a nearby ﬂower and a distance mountain sweeps from under the photographer’s feet passing through both. The rules may still be used — but not for close up work.
and substantially with base tilt cameras. and T B is the tilted back plane. it is also necessary to tilt the lens to bring the subject into focus.5 Back Tilts Tilting the lens preserves perspective because the magniﬁcation deﬁned by v/u remains unchanged.mm apart. If h and y are distances as shown in ﬁgure (3). and thus tilting leaves magniﬁcation unchanged. It moves slightly with center tilt cameras. and the perspective. tilting the back changes the v/u ratio.4 Lens Movement Figure (6) places the center of the lens at the same point for both a tilted and untilted lens. When the back is tilted. For a center tilt camera. Parallel lines can be made to meet at vanishing points and vice versa. hence usually in front of a. and the desired lens tilt is given by α + β. then Wheeler’s tilt rule indicates that lens should be tilted forward until the top of the standard has moved forward 25*12/60=5 cm. sin(γ − α) t sin(γ − α) m + 1 sin γ (13) = arcsin . Wheeler’s rule of 60 indicates that the lens should be tilted by 60*12/60=12 degrees. since F P . the F P for a tilted lens will be at approximately K − ∆. where LP and F P are the untilted lens and focal planes. But in fact the lens center is not at the same point for the two situations. and J is in meters. which is larger than K = uf /(u − f ) for the untilted lens. If the height of the front standard measured from the tilt axis to the top of the standard is 25 cm. for a tilted lens. where t = H/V in ﬁgure (7). which is usually a negligible change from φ = arcsin(f /J) because ∆ is seldom more than about 50 millimeters. the F P will usually be in back of a. This has little eﬀect on the calculation of φ since if the lens center in ﬁgure (6) is moved horizontally by ∆. then y/h = (y cos ω)/(h cos ω) = v/u for any tilt. is at a horizontal distance of uf /(u cos(φ) − f ) from the lens center. 5. 5. Lens movement has more eﬀect on the focus position. The back is tilted α degrees. Figure (7) shows the relevant planes. For a base tilt camera. and m is the magniﬁcation as in ﬁgure (6). one has φ = arcsin(f /(J − θ)). It may be seen that D= and thus β is given by β = arcsin f sin(γ − α) J sin γ J sin γ . On the other hand. 14 . The lens moves as it is tilted. where θ = ∆/ tan(γ).
DOF. and FP are parallel. hence the near DOF limit Zn = vdf uf 2 u(H − f ) = 2 = . 15 (15) . one sees that x = cv/(d − c). LP.1 Near DOF limit For ﬁgure 8. vd + f (d − c) f + (u − f )N c H + u − 2f f2 +f Nc (14) where N = f /d is the fnumber. (v + x) − f is By similar triangles. equations for the case when SP.SP TL TLP α LP β TB FP α V H D f γ J Figure 7: Lens tilt for tilted back 6 Depth of ﬁeld for parallel planes Figures 8 and 9 show diagrams that will be used in deriving depth of ﬁeld. The near DOF distance is given by equation (3) as Zn = (v + x)f . 6. and H= is the hyperfocal distance. IP is behind FP. the circle of confusion arises because the image plane. respectively. The diameters of the aperture and circle of confusion are d and c.
then one has Zn = uHa . It follows that the near distance from the u(t)f 2 u(t)f u(t)f + (u(t) − f )N c) +u(t)f /(u(t)−f ) = . one has t = δ + u2 /(u − f ) 1 2 + 1 4 − f t−δ . Since t = u + v + δ. from the focal plane. f2 If one takes Ha = N c .LP FP IP d/2 c/2 x v Figure 8: DOF near LP IP FP d/2 y c/2 v Figure 9: DOF far The Front DOF is u − Zn = u(u − f )/(H + u − 2f ). t. + (u(t) − f )N c u(t) − f f 2 + (u(t) − f )N c or perhaps more succinctly 16 . where δ is the nodal space. Ha + (u − f ) (16) and u(t) = (t − δ) focal plane is Znt −δ = f2 One may also recast these equations in terms of the distance.
so DOF = 2u(u−f )(H−f = (H−f )2 −(u−f ))2 ∞ for u < H for u ≥ H . f2 If one takes Ha = N c . from the focal one has for u(t) < H Zf t −δ = u(t)f 2 u(t)f u(t)f − (u(t) − f )N c) +u(t)f /(u(t)−f ) = . hence the far DOF limit is Zf = uf 2 f 2 −(u−f )N c = u(H−f ) H−u ∞ for u < H for u ≥ H (17) The Back DOF is Zf −u = u(u−f )/(H −u) for u < H and inﬁnity otherwise. 6.Znt −δ = Zn u(t) N + (u(t) − f ) f = Zn t Nc + f (1 + 1/m) f = Zn t f + f (1 + 1/m) Ha .2 Far DOF limit (v − y)f . Ha − (u − f ) (18) In terms of the distance. one has y = cv/(d + c). 6. the denominator vanishes because the rays from the object are parallel. (v − y) − f For ﬁgure 9. (19) 17 .3 DOF 2N c(u−f )uf 2 f 4 −[N c(u−f)]2 The DOF is the diﬀerence Zf − Zn . vd − f (d − c) f − (u − f )N c When y is zero. − (u(t) − f )N c u(t) − f f 2 − (u(t) − f )N c u(t) Nc − (u(t) − f ) f t Nc − f (1 + 1/m) f t f − f (1 + 1/m) Ha f2 or perhaps more succinctly Zf t −δ = Zf = Zf = Zf . giving u = By similar triangles. IP is in front of FP. and u = vdf uf 2 = 2 . t. then for (u − f ) < Ha one has Zf = uHa . where m is magniﬁcation.
one should use DOF = 2N c(m + 1) .or DOF = 2N cvf (v−f )2 −(N c)2 2vf (H−f = (v−f )2 (H−f )2)−f 4 ∞ 3 for v < Hm for v ≥ Hm . In general. and I the image height. The near and far limits in terms of m are: Zn = and Zf = f (1+m) m− N c f f (1 + m) . An interesting variant of this may be obtained by substituting equation (6) into equation (19): DOF = 2N c m+1 m2 1−K ∞ for K < 1. for K ≥ 1 (20) Nc where K = ( mf )2 . m2 − ( N c )2 f (22) This equation shows how DOF decreases as f increases. which corresponds to about 290 mm at 25 cm. and viewing such a print at 25 cm completely ﬁlls the ﬁeld of view. It is interesting to note that the denominator of equation 22 vanishes when u equals the hyperfocal distance as given by equation 15.4 Circles of confusion The human eye has a resolution limit of about 5 line pairs per millimeter (l/mm) at the nearest distance of distinct vision (about 25 cm for a normal eye). which for small K (as might occur in macro photography) becomes 2 m+1 S S . The diagonal of an 8x10 print is 325 mm. m + Nc f (23) for mf > N c for mf ≤ N c ∞ (24) 6. say. In practice. One can check this for oneself with a lens chart. This means that a normal human eye can just discriminate between ﬁnely drawn lines separated by 0. It follows that 5 l/mm is the maximum resolution needed by a print. this means that DOF depends on f when u > H/2. the perspective will be correct if the photograph was taken with 18 . The ﬁeld of view of human vision is about 60 degrees. (21) DOF ≈ 2N c = 2N c + m2 I I where S is the subject height.2 mm. but not for closer subjects. For such a print viewed at such a distance.
setting it properly means that those objects at the DOF limits will just be resolvable in a print. u. To a large extent this opinion derives from a selective modiﬁcation of the parameters: leaving some ﬁxed and changing others.025/mm. then the DOF decreases as the format increases.a standard lens3 . with m = 1 and the other parameters as above. 4 for 4x5. is an important variable. and f = 50 mm.025 mm. gives the spacing between lines. the diameter of this circle. 6. c x y 3 One who’s diameter is approximately equal to its focal length 19 . c. The x axis indicates the format in terms of the size of the short side of the ﬁlm — 1 for 35 mm. One concludes from this that DOF decreases with format size when the fnumber is ﬁxed. This means that to preserve perspective. but remains unchanged or slightly increases when the fnumber number is scaled with the other parameters. one should view a 16/20 print at 50 cm. One has u = 100 mm for the 35 mm format and u = 400 mm for the 4x5 format. For a 35 mm camera. For DOF.6 The DOF equation An interesting equation similar to the Gaussian lens equation may be derived for the near and far DOF limits. This circle is called the “circle of confusion.025 mm. but this is not so if the fnumber is scaled as the other parameters are scaled.” The circle of confusion thus depends on the degree of magniﬁcation of the image with respect to the size of a print taken with a standard lens. As may be seen throughout these notes. one should have at least 5 l/mm. 6.01. magniﬁcation comes into play: thus there will be a factor of 8 for a 35 mm image. For an 8x10 contact print. c = 0. The lens to subject distance. and other prints proportionally. it should be at most about 0. This implies geometrically that the ﬁlm resolution should be eight times 5 l/mm or 40 l/mm. 0. Figure (10) on the right shows the DOF in the ﬁxed and scaled cases — note that the DOF scales are in millimeters on this graph. If one ﬁxes the fnumber. When the image size is smaller. The reciprocal of this. For macro photography. From ﬁgures 8 and 9 one sees v+x v−y d = = = r. is approximately constant at 5 m for the x’s shown. m = 0. and should decrease as the print size increases past 8x10. and the diameter of a circle at the limit of human discrimination.5 DOF and format There is always heated argument about the DOF achieved from various formats. The general view is that larger formats have smaller DOF. Figure (10) on the left shows the DOF when N = 16.
v r 1r+1 .7 Hyperfocal distance The hyperfocal distance in equation (15) is the distance at which the far limit of DOF becomes inﬁnite. u v = = 1r−1 . Substituting this as u into the formula for the near limit. v r A more straightforward derivation is to sum the reciprocals of equations (14) and (17).Figure 10: DOF for Fixed and scaled N thus 1 v+x 1 v−y and since 1 1 + Zn v+x 1 1 + Zf v−y one has by addition 1 2 2 1 1 + = + − Zn Zf u v v or the DOF equation 1 2 1 + = . Zn Zf u (25) r−1 r+1 + r r . u v 1 1 + . equation (14) gives 20 . 6. = = 1 1 + .
As u increases. Thus one may focus on inﬁnity. 6.1 Focus given near and far limits Object distances An interesting question. 2 thus focusing at the hyperfocal distance puts everything in focus from half this f2 f distance to inﬁnity. It is hard to imagine situations in which these small diﬀerences will matter. the hyperfocal focus is best determined from distances on the image side of the lens. and then move the standard by N c to obtain focus at the hyperfocal distance. is “what is the focus distance. and Zf (1 − N c/f ) ≈ Zf . and that DOF ≈ DOF + (Zn + Zf )N c/f . giving the usually cited formulas H Zn ≈ = f2 . As a practical matter. Zn in equation (14) may be seen to approach f 2 /(N c). A calculation shows that Zn (1 − N c/f ) ≈ Zn . Simple substitution into equation (15) shows vH = f + N c = f (1 + Nc ).9 6. a good approximation to H.9. f 2 + uN c H +u (27) (28) Zf = uf 2 f 2 −uN c = ∞ H u H −u for u < H for u ≥ H for u < H for u ≥ H (29) DOF = 2N cu2 f 2 f 4 −(N cu)2 = ∞ 2H u2 H 2 −u2 (30) The approximate DOF is slightly larger than the exact DOF. Nc uf 2 Hu = . given the near and far DOF limits?” 21 .Zn = H . Another often quoted deﬁnition for “hyperfocal distance” is the value of Zn when u = ∞. 6.8 Approximations These equations simplify by noting that in many cases u − f ≈ u. f (26) where vH is the position of the standard when focused on the hyperfocal distance. It is worth noting that H = Nc + f = f (1 + Nc ).
and that 2x. u f θ where θ is the harmonic mean of zn and zf : θ= 2zn zf . For u one has 1 1 1 = − . say.2 Image distances The expressions simplify considerably when one uses the image distances zn and zf corresponding to Zn and Zf respectively. = −1 + 1+ 8αx α − 1 N c (α + 1)2 Thus from two distances x and αx.The DOF Equation. nontrivially. with α > 1 one may obtain the focus distance and either the fnumber or the focal length that will make these two distances the near and far DOF limits.9. corresponds to focus at the hyperfocal value. Zn Zf u which when solved for u gives u= 2Zn Zf 2αx . 2 ± 1 . 4 “The focus point is one third of the distance between the near and far limits. is 1 1 2 + = . for x > 0. and that the commonly stated one third rule4 is correct. equation(25). 2 One may solve for N given f in equation (14). only when α = 2. 6.” 22 . = Zn + Zf 1+α (31) where x = Zn and α = Zf /Zn . It is useful to note that equation (31) ranges between 1x and 2x. It is also interesting that approximations for f and N may be obtained by equating the focus distance to the hyperfocal distance. although many will be satisﬁed when using it in the range of. Note also that u is the product of the limits divided by their average. obtaining N= Zf − Zn f2 α−1 f2 = . zn + zf (33) Equation (33) may also be obtained from equation (2). . c (2Zf Zn − f (Zf + Zn )) c (α(2x − f ) − f ) (32) and for f given N : f = Zf + Zn 1 Nc 2 Zf − Zn 1 α+1 Nc 2 α−1 −1 + 1+ 8Zf Zn Zf − Zn N c (Zf + Zn )2 .
u f µ For N given f one may write from equation (32) cN = ρ ρ f = . The expression for f given N is of course f= µcN . This gives DOF = Zf − Zn = 2f 2 ρ 2f 2 ρ = . ρ 7 Depth of ﬁeld. µ 1+m (34) where ρ = (zn − zf )/2. Thus (v − ρ) and (v + ρ) are the conjugates of Zf and Zn . (v − f )2 − ρ2 (mf )2 − ρ2 (35) Equation (34) may be used to ﬁnd m = (ρ − N c)/N c. which is likely that used by Sinar for their DOF calculation. and since all three means diﬀer by at most about 1%. ρ= DOF f (36) + + Nc f 2 . one can write equation (33) as 1 1 1 ≈ − . 23 .Since θ = λ/µ. which then gives ρ as a function of DOF: 2 2 2 Nc N cf Nc . When the magniﬁcation is small. depth of focus The function relating DOF and ρ may be obtained by assuming the depth of focus range 2ρ is equally divided by the focus point v. one has the approximation cN ≈ ρ. and m is magniﬁcation. where λ is the square of the geometric mean and µ the arithmetic mean of zn and zf . 1 + Nc + 1+ ρ= 2 − 1− f − (N c)2 DOF DOF f or N cf 2 1 + Nc ρ= 2 2 f − (N c) DOF Nc + 2 DOF Nc DOF 2 from which one may express ρ as a function of DOF and m: 2 mDOF f2 −1 + 1 + .
of p. From similar triangles (d − r)/(t − r) = (v + x)/x one ﬁnds t= 5A rv + xd . P. then p would equal c. which will produce an image with diameter k times the diameter of a circle corresponding to a point image at that distance. and objects made up of such circles are fuzzy. The diameter of the aperture is d The diameter of the image of an object point is shown as p. and its image plane. thus r=S f v+x =S . u u −f from equation (3). Knowing the degree of fuzziness for an object can be useful. SP u P u' LP v FP x IP S d p r t Figure 11: Object in front of SP The magniﬁcation of the object with respect to IP is r/S = (v + x)/u . The problem is to calculate S corresponding to a multiple. If P were the near DOF plane.8 Fuzzy images Points image as circles on the ﬁlm. Points in other regions. IP with a circular5 object of diameter S on P. These are those between the DOF limits. although still fuzzy. and t on FP. while one with k = 10 will probably be recognizable. The image of this object has diameter r on IP. as when one wants letters on a sign in the background to be unreadable. v+x ﬂat disk without depth. Figure 11 shows the three planes as well as a plane. say k. This may be done by calculating the size of an object at a given distance. Some of the image circles are so small that the eye cannot distinguish them on a print or slide. image as circles which the eye can distinguish. Thus an object with k = 2 will be quite fuzzy and impossible to recognize. 24 .
The argument is essentially the same as above. one ﬁnds x=v+x−v = uf uf f 2 (u − u ) − = . then t= thus (k − 1)xd = rv = r and using equation (37). It is worth noting that the above derivation holds if S is not symmetrical about the lens axis. S = (k−1)d xd rv + xd = kp = k . u−f u −f u−f f u−u xf 2 (u − f )(u − f ) x = (k−1) = (k−1) . At the near DOF limit. v−y and when P is the far DOF limit: 25 . t= The equation is S = (k − 1) yf 2 f u −u = (k − 1) . v+x v+x f uf uf =S . this becomes S = (k − 1) f N 1− Zn u . t−r y thus. and one obtains p= Let t be a multiple of p. (38) f 2u N u N v(v + x − f ) with N = f /d. u −f u−f (u − f )(u − f ) xd . The inset shows the triangles.Again using equation (3). which give d+t v = . N u N v(f + y − v) (39) rv + dy . The only small problem is ﬁnding the similar triangles involving t. v+x (37) From similar triangles p/d = x/(v + x). Figure 12 shows the planes with P to the left of SP. say t = kp.
u The derivation does not depend on the symmetry of S about the lens axis. The disk 26 . u (u − f ) u (u − f ) u (u − f ) u − u d t− .P SP u u' LP v IP FP y S d r pt (d+t)/2 (tr)/2 LP IP FP Figure 12: Object behind SP S = (k − 1) f N Zf −1 .1 Eﬀects of diﬀraction on DOF Theory Light from a point source. This is due to the fact that the path of light near an edge is longer than that away from the edge causing a phase diﬀerence. Lord Rayleigh concluded that two equally bright point sources could just be resolved by an optical system if the central maximum of the diﬀraction pattern of one source coincided with the ﬁrst minimum of the other. fu u t S = = 9 9. and complete interference occurs at a certain distance from the center. diﬀracted by a circular opening. and at harmonic distances further away. From a study of the appearance of the diﬀraction patterns of closely spaced points. is focused by a lens not as a geometrical point. Useful expressions for t and S in terms of each other are: f u − u d fu + S. but as a disk of ﬁnite radius surrounded by dark and bright rings.
It follows that for human vision.2 mm or 0. one has: s = 1. as the frequency of two just separable lines. The angle α from the center of the lens subtended by two points with this spacing is given by sin(α) = (1. Since N = f /D. 35. where f is focal length. l/mm. Kλ n (41) The spacial frequency 1/s gives lines per millimeter. The diameter of the diﬀraction circle of confusion.008 in.1 which when substituted in equation (40) gives N ≈ 25 mm. about the point of focus. p260. thus a spacial cycle consists of a black bar and a space. and n the index of refraction. Kλ varies from about 1000 1 to about 2000 1 . This may be conﬁrmed by inspecting an Edmund USAF pattern. The 35 observations in this table were obtained by using a 150 mm Sironar lens with Velvia. δ mm 7 . and f ≈ 25 mm. where n is the index of refraction on the image side of the lens.2 Data Diﬀraction aﬀects resolution so that the parameter c used in equations. placing the contrast ratio. D is the diameter of the aperture. but in the ball park (In bright light the iris can close down to about 2 mm. and N is the fnumber. The contrast between the dark lines and the white background was approximately 5 stops. l/mm. with s/2 the common width. or simply “lines per millimeter”. Test charts allow the selection of “just separable” pairs of lines. These correspond to the spacial frequency 1/s. is s. one ﬁnds 1 1 the object distance to be s = sin(α)u. this gives m = 0. such as equation (34). approximately midway 6 records 7 Note: 153157 δ = 2ρ in equation (34) 27 . s ≈ 1/5 = 0. lp/mm.). m Kλ n (40) Since visible light ranges from about 400 × 10− 6 mm to about 700 × 10− 6 mm. It is generally agreed that normal human vision can just resolve about 5 l/mm on a test chart at the minimum distance of distinct vision (about 250mm). Assuming the human eyeball is about 25 mm. Table 1 6 shows resolution in l/mm as a function of N and depth of focus. 9. m is magniﬁcation. where λ is the wavelength of light. black bars are used separated by spaces of the same width. called either “line pairs per millimeter”. This is s ≈ N (1 + m) mm. c. For test charts.deﬁned by this ﬁrst minimum is called an Airy disk. Often one takes (mm) (mm) Kλ = 1500 1 for a medium wavelength λ = 550 × 10− 6 mm: see Sears (mm) (1958). which is a bit small. For an object at distance u. this gives D ≈ 1mm. The distance s on the other side of the lens is given by s = ms(n/n ).22N (1 + 1 N (1 + m ) 1 )λ/n ≈ mm. Since u/D = f (1 + m )/D = N (1 + m ). changes as other parameters change.22λ)/(nD).
1.25 Table 1: Raw resolution data for Velvia fnumber 64 45 32 22 16 11 8 0 22 28 28 31 28 2.5 4. corresponding to a spacial frequency of 10 l/mm.2 2. but at some small value.6 and 100 contrast ratios.7 Table 2: Raw resolution data for TMax 100 between the usual 1. For normal lens’s where the focal length is approximately equal to the ﬁlm size.4 6.1 17 8 3.6 it will be seen that the data for δ = 0 does not agree with equation (43).3 δ mm 5 7.fnumber 64 45 32 22 16 11 8 0 23 23 26 32 32 23 26 2. Table 2 8 shows a set of data taken with TMAX 100 using a 210mm Sironar lens — all other parameters are the same as for Velvia. In Section 9. and so the diameter of the onﬁlm circle of confusion should be one half s or 0. It seems likely that the focus when taking the data was not at δ = 0. The residuals had a sample standard deviation of 3. The locus of equation (34) is marked on ﬁgure 13 by the line labeled c = 0.5 20 20 26 23 23 16 14 9 9 6.5 23 23 32 26 18 18 9 δ mm 5 7. Lines for c = 0. which is of the order of experimental error. since one can read the USAF charts only to the nearest 2 l/mm.5 20 25 22 11 4.5 2. one should view a picture at a distance consistent with that of the taking lens. this means that an 8x10 will be at the distance of distinct vision.4 4.5 3. The c = 0.4 4. The data was smoothed by a low order polynomial and is shown as Figure 13. This suggests that the data may be in error.2 10 17 11 9 3. For a 4x5 ﬁlm. this means that one views a picture x inches high from x inches.1 l/mm.1 lines seems to be 8 record 149 28 .1 3.05 and c = 0.1 mm. To preserve perspective.1 1.2 are also shown.9 2.5 20 20 20 14 6. The USAF target was 1500 mm distant from the lens giving a magniﬁcation of 1/9.5 10 20 23 10 6. perhaps δ = 0.5 or larger.
N increases to a maximum and then decreases. with the front and back DOF being 130 mm and 160 mm. however.05 is 290 mm.64 FSTOP c=0. This ranges from bill number 5 through bill number 12. 9. and inversely for diﬀraction. In ﬁgure 17 the dollars were arranged 30 mm apart on a slanting board. an array of dollar bills was photographed and the ﬁlm examined under magniﬁcation. but the data has a standard deviation of at least 2 l/mm. one would need a ﬁlm resolution of 20 l/mm. The line c = 0. A Pentax 645 camera with a 150 mm lens at f32 was positioned 1500 mm from the central pair of bills (numbers 8 and 9. Both the data and the choice of ﬁtting model could be improved upon. increases directly with N for depth of focus. The results are good enough.05 shown in ﬁgure 13 is slightly above the 20 l/mm contour.05 for a 645 format seems reasonable.1 28 32 16 8 6 18 1 12 0 1 20 24 22 c=0. marked with dots on their corners).2 6 4 2 4 6 8 10 Figure 13: Resolution as a function of N and δ for m=1/9 somewhat above the 10 l/mm contour. This is due to the fact that lines per millimeter.3 Resolution It is useful to observe how the contours change in ﬁgure 13. The DOF using c = 0. to show that equation (34) is reasonable. For large δ. Noticeable degredation seems to occur in the neighborhood of bills 5 and 12. As an additional check. For small δ’s. An 8 power loupe was used. hence a 9 power loupe will bring the image to full size.05. however. which corresponds to c = 0. the lines per millimeter increase steadily with N because of the shrinking aperture. Since 4x5 prints are often printed at 16x20.05 24 c=0. so the choice of c = 0. Figure 15 indicates the eﬀect of decreasing 0 δ mm 29 . The magniﬁcation is 1/9 = 150/(1500 − 150).
64 FSTOP Opt2 24 20 Opt Hansma 28 32 16 8 6 18 1 12 0 1 24 22 6 4 2 4 6 8 10 Figure 14: Optimum N as a function of δ for m=1/9 the aperture on the image circle. For depth of focus. The eﬀect of diﬀraction on resolution is just the opposite. 1+m (42) which is the maximum N for a given δ. decreasing the aperture is seen to decrease the image circle. giving N= Kλ δ/2 . equation (34) gives Lδ = (2N (1 + m))/δ. since decreasing the aperture makes the image more diﬀuse and decreases the resolution. thus increasing the resolution. one has Ld = Kλ /[N (1 + m))]. since the product Ld Lδ = 2Kλ /δ does not depend on N . The point at which the diameter of the depth of focus circle of confusion equals the diameter of the diﬀraction circle of confusion is obtained by equating Lδ to Ld . From equation (41)9 for diﬀraction. For a ﬁxed depth of focus with a constant distance between IP and FP. 9 Equation (40) is not appropriate since here one needs to compare resolutions on the image. This ﬁgure may be compared with ﬁgures 8 and 9. 0 δ mm 30 .
DOF. Figure (16) shows equation (42) for this 35 mm scaling.5. From Section 6. He seems to have made several errors such as using 2s for the diﬀraction circle of confusion. Hansma found the optimum N to be N = (375δ) for m = 0. It is important to note that these changes do not eﬀect DOF. ignoring the magniﬁcation factor in equation (40). Equation (42) seems to agree better with the observed optimum than does Hansma’s. Although the resolution values will diﬀer among formats. When the far limit is inﬁnite. and even confusing equations (40) and (41). In any case. Equation (42) is marked “Opt” in ﬁgure (14). 9. The diﬀraction resolution Ld is aﬀected in a very interesting way by format. The limit of equation (36) as DOF goes to inﬁnity is δ = 2mf . This is 1/ (2) of equation (42). the near limit is half the hyperfocal distance. and this works out to δ ≈ 2mf where m is the magniﬁcation for an object at half the hyperfocal distance. Equation (36) gives ρ = δ/2 as a function of magniﬁcation. This assumes depth of focus on both sides of the focus. The apparent intersection of the contours with the δ = 0 vertical are likely to be an artifact the polynomial form and measurement error. The curve “Opt2” represents 2Lδ = Ld . For a ﬁxed object size at a ﬁxed distance from the lens.LP FP IP d/2 c/2 x v Figure 15: Eﬀect of aperture on the image circle. He chose an empirical combination of depth of focus and diﬀraction by taking the RMS of the two. m/3 and ρ/9. Hansma (1996) attempted to combine depth of focus and diﬀraction. as may be seen from equation (35) by substituting f /3.4 Format considerations It is worth noting that equation (42) does not depend on format. reducing mf by about 1/9. since any linear combination of powers of the two circle of confusion expressions will produce the same optimum. the relationship between N and δ does not change. The particular choice of RMS has no eﬀect. the m and f for 35 mm are both about 1/3 that for 4x5. and f . N must change as format changes in order to keep DOF 31 . m. which may help to visualize the eﬀect of diﬀraction on resolution.
and any substantial increase in N will decrease resolution.5 Minimum aperture The curve from equation (34) using the appropriate c lies well below the Opt curve in ﬁgures (14) and (16). is about 1. Lens manufacturers use this to choose the minimum fnumber for their lenses. Since this is smaller for smaller formats. hence resolution is a function of the distance between lens and ﬁlm (f (1 + m)). hence smaller formats must have greater resolution.025 0. δ) comes close to the optimum. The δ at which the Opt curve meets the c curve is 2Kλ × c2 . and they do! Another way to view this is to examine equation (41) with N = f /A.4 0. the smallest aperture that is practical is about 2 mm. 0 32 .8 1. hence 35 mm images must produce about 3 times the resolution of 4x5 images. At such extremes the point (N. which for Kλ = 1500. δ) is usually well away from the optimum curve. their resolutions must be greater. 9.2 0. For any format. The change from 4x5 to 35 mm is about 1/3. except at extreme depth of focus limits for the format.64 FSTOP Opt2 Opt 32 16 8 c=0.8 for 35 mm and 30 for 4x5 — well outside any practical range. where A is the aperture. For any format the point (N.6 0.0 δ mm Figure 16: Optimum N as a function of δ for 35 mm constant: but Ld is inversely proportional to N (1 + m).
(43) Kλ δ/2 . The data at δ = 0 was not used to make the contour plot. and may be compared with ﬁgure (13). suggesting that it captures a fundamental aspect of the problem. except for δ = 0. One choice is (1/Lα )α = (1/Lδ )α + (1/Ld )α . Values calculated from Lα agree reasonably well with ﬁgure (14) in view of the 2 l/mm error in the data. which may be compared with tables 1 and 2.6 Theoretical curves One must combine Ld and Lδ in order to ﬁnd the actual resolution. which gives Lα = Kλ N (1 + m) ((Kλ δ/2)α + (N (1 + m))2α ) 1/α .Figure 17: DOF determination 9. The predictions for the 35 mm case produce resolutions of a magnitude sim 33 . A contour plot of this data is shown in ﬁgure (18). Substituting equation (42) into this gives the maximum Lα = 2−1/α The agreement between equation (43) and the data is good. The combination should approach one of the resolutions as the other becomes extreme. Table 3 shows values calculated from equation (43). thus the plot does not show the singularity at δ = 0.
7 168.2 10 11.9 4 19.0 42.9 4.8 1 20.7 27.4 84.2 32.3 38.2 22.1 4. 34 .1 4. fnumber 64 45 32 22 16 11 8 0 21.4 19. 64 FSTOP 32 35 30 25 20 16 8 15 5 10 3 0 2 4 δ mm 6 8 10 Figure 18: Contours for theoretical resolution from eq (43) with α = 2.7 6.2 25. This is consistent with the conclusion in Section 9.0 8 13.5 8.ilar to what one expects from 35 mm.8 6.4 1.2 61.7 36.8 24.2 16.0 4.4 12.2 8.8 14.7 δ mm 2 20.9 3.4 3.0 8.4.6 11.4 12.4 8. The implication of this is that diﬀerences in the formats are principally the result of size and not of superior lens quality for smaller formats! Both lens quality and focus precision play a role.2 2.8 9.0 17.9 28.4 6 15.8 Table 3: Theoretical resolution from eq (43) with α = 2.1 5.7 17.6 2.1 3.5 7.6 11. but the higher resolutions for small format appear to be due to operating in the range of smaller δ’s.1 30.4 122.
2 Near and far DOF equations in terms of ρ It is often more useful to calculate these limits in terms of ρ instead of h. f 2 + [h cos(φ + ρ) − f ]N c and the near DOF along the ray is Zn = u / cos(φ + ρ). LP. then u = h cos(φ + ρ). The image on FP is an ellipse.1 Depth of ﬁeld for a tilted lens Near and far DOF equations The DOF equations above hold for tilted lenses with minor modiﬁcations. There is one minor problem: the diameter of the circle of confusion is not correct. The above assumes the image is on I not FP. 10. Figure 19 shows the three planes SP. and its maximum diameter varies slightly with ρ. Planes F and I are parallel to LP and have been drawn through the intersections of the ray with SP and FP. The distances of F and I from LP are u and v. The far DOF limit is derived in the same way and is Zf (ρ) ≈ hf 2 f 2 −[h cos(φ+ρ)−f ]g ∞ for h < H(ρ). To this end. Similar calculations follow through when the ray is below the lens axis. Equation (14) may thus be used to ﬁnd the near DOF which is labeled u in ﬁgure 19. it may be seen from ﬁgure (20) that h cos ρ = J + h sin ρ. and FP with a ray at angle ρ. shown as NP in ﬁgure 19. and passes through the pivot point.10 10. f 2 + [h cos(φ + ρ) − f ]g (44) where g ≈ N c. Let h be the distance from the lens center L along the ray to SP. is a straight line. respectively. The graph of this equation. result in equation (44) with ρ a negative quantity. tan γ hence 35 . for h ≥ H(ρ) (45) where H(ρ) = f 2 /(g cos(φ + ρ)). thus equation (14) becomes u = hf 2 cos(φ + ρ) . with constant g. It corresponds to the image plane IP behind FP. and following the sign convention used above. thus the near DOF for a ray at angle ρ is: Zn (ρ) = hf 2 .
cos(ρ + γ) f 2 sin γ . f sin φ cos(ρ + γ) + g cos ρ sin(γ + φ) (46) Zf (ρ) = f 2 sin γ f sin φ cos(ρ+γ)−g cos ρ sin(γ+φ) ∞ for nonnegative denominator for negative denominator (47) 36 .NP LP u u' VP FP IP SP ρ+φ φ Zn ρ h L v v' F I Figure 19: Tilted DOF diagram h= and thus. using J = f / sin φ. Zn (ρ) = and J sin γ .
1 2 1 + = .1 Focus given near and far limits Object distances Just as for parallel planes.3. holds for Zn (ρ) and Zf (ρ). An equation similar to the DOF Equation. one may obtain the focus distance from the near and far limits. Zn (ρ) Zf (ρ) h (48) which is most easily seen by assuming g constant and summing the reciprocals of equations (44) and (45). The DOF for a given ray angle ρ is DOF (ρ) = 2g(h cos(φ + ρ) − f )hf 2 . f 4 − [g(h cos(φ + ρ) − f ]2 An expression for N given f is 37 .3 10. equation (25).LP SP a h FP ρ b L J Figure 20: Finding h from ρ 10.
Zn (ρ) cos ρ The tangent of the graph of Zn (ρ) is T = tan ρ + On substituting Zn (ρ) as given by equation (46) into this. one has µ cos(φ) = f (1 + m).3. and one may choose to measure δ and µ along the original lens axis. Since the parameters c.N= f2 Zf (ρ) − Zn (ρ) . one has cN = δ cos(φ) . with ρ = 0. 10. which is equivalent for practical purposes.9. c 2Zn (ρ)Zf (ρ) cos(φ + ρ) − f (Zn (ρ) + Zf (ρ)) 2Zn (ρ)Zf (ρ) cos(φ + ρ) − f Zn (ρ) + Zf (ρ)) . the ratio δ/µ is also constant for all ρ. The expression for f given N is of course f= µcN . . N . Zf (ρ) − Zn (ρ) (49) and for f given N : f= g One might use these expressions in calculating the lens tilt by replacing the distance to SP along one of the rays with the distances of the desired near and far limits. since φ is never very large. one has T = 1 g + tan γ f 1 1 + tan φ tan γ 38 . and remembering that m is a constant for all angles. with zn (ρ) and zf (ρ) the conjugates of Zn (ρ) and Zf (ρ) via equation (2). 1+m (50) This may be compared with equation (32) in section 6.2 Image distances Using equation (2) in the form hf /(h cos(φ + ρ) − f ) in equation (49) gives f cN = δ . 2δ cos(φ) 10.2. and using J = f / sin φ. In this case. µ where δ = (zn (ρ) − zf (ρ))/2 and µ = (zn (ρ) + zf (ρ))/2. and f are constant in this formula.4 The graph of the DOF limits J .
with γ− the angle of the far DOF plane with respect to the vertical. the size of δ is constant because x/j = δ/(d − δ) is constant — circles of confusion remain circles and do not turn into ellipses. j. The fall for Zf (ρ) is also approximately equal to J. (d − δ) cos(ρ + φ) δ f sin γ . save for a negative sign in front of g. and thus must ﬁnd δ. and hence if g is constant. then one may write and I are parallel. Let H = f 2 /g. Similar comments apply to Zf (ρ): the equation is the same. hence the ray divides δ into two equal parts. d − δ cos ρ sin(γ − φ) δ 2 cos φ. 39 .5 Variation of maximum diameter of the ellipse of confusion with ρ We will show that the variation of the maximum diameter of the ellipse of confusion is small. 11 Note: g = Nδ. Even more. where γ+ is the angle of the near DOF plane with 2 respect to the vertical. tan γ This is practically diﬀerent from J. then one sees from T. and . only for very small γ or large φ. Since arctan(T ) is π − γ+ .which shows that the graph depends on ρ only through g. The corresponding image on I. one has 1 tan(γ+ ) 1 tan(γ− ) = = 1 tan γ 1 tan γ + − g f g f 1 tan φ 1 tan φ + + 1 tan γ 1 tan γ . An interesting observation is that the rise of the Zn (ρ) graph above SP at the hyperfocal distance is approximately equal to J. and consequently the DOF limits are approximately linear. 10. On FP. Figure 21 is a modiﬁed portion of ﬁgure 19. We desire c = c1+c2 to be constant. has diameter11 δ. centered10 about the slanted ray. the image is an ellipse with extents c1 above and c2 below the ray. and note that H/ tan γ is on SP. along the ray from LP to FP is v/ cos(ρ + φ). above that the rise above SP is J cos φ + f . and similar triangles give δ/x = d/(j + x). the graph is a straight line. The distance. or x= or after some manipulation: x= Let p = 10 LP δv . The aperture is the slightly darkened portion of LP with diameter d.
γ = 80◦ . The parameters are φ = 5◦ . Figure 22 shows a comparison of the near DOF plane. a = x sin ρ. becomes inﬁnite when ρ ≈ 22. and N = 32. with a graph of equation (46) using g = N δ obtained by solving equation (51) for δ as a function of c.6◦ : the 40 . and b = x cos ρ. The dotted curve. hence one may reasonably take g = N c as a constant in equations (46) and (47). b+e p−a . representing the graph of equation (46). Adding and simplifying gives δ ( 2 )2 cos φ sin φ − x2 sin ρ cos ρ δ (x cos ρ)2 − ( 2 sin ρ)2 c = δ cos φ + δ sin φ (51) and for paraxial rays. the expression on the right diﬀers negligibly from δ. c2 = p + e b−e δ where e = 2 sin φ. F = 150mm.LP FP φ e b c1 IP d/2 φ ρ L ρ+φ d/2 j v δ/2 p c2 δ/2 e x ρ a p I Figure 21: Ellipse of confusion diameter for tilted lenses c1 = p − e p+a .
then the ratio L/Hf increases with N 2 /t. In fact if Hm is measured 41 . It turns out that the constant of proportionality can be obtained by a physical argument. and the luminance is ﬁxed by the scene. This is an extreme example. L/Hf is proportional to N 2 /t. which are the fnumber. and increased by increasing t. at the focal plane.5º 10º LP 5º 0º .angle of the near DOF plane. required to produce a density of 0. with the constant of proportionality being ﬁxed by the scene and the ﬁlm density desired. so the only thing remaining is to specify Hf in terms of a ﬁlm.1 over background and fog. of the scene and the exposure. 11 11. Hf . and (2) the value of Hf that will produce the desired density on the ﬁlm.1 Image control Exposure There are two quantities of prime interest: the luminance. In fact. N and the exposure time t. the photographer needs only two things to make an appropriate exposure: (1) the constant of proportionality.15º FP L SP 6 5 0º 4 º 8 5º 2000mm 1500mm 1000mm 500mm 0mm Figure 22: Comparison of near DOF plane (solid) with true curve (dotted). NP 7. The luminance is modiﬁed into the exposure by the camera controls. the diﬀerences are usually smaller. Hm . The exposure is decreased by increasing N . Now black and white ﬁlms are characterized by the exposure. This means that if one ﬁxes the scene luminance. Since N and t are under the control of the photographer. L. Thus Hf is made smaller by increasing the ratio N 2 /t.
1 density.in cd/m2 . Many users ﬁnd the calibration used by the Minolta Spotmeter F more appropriate for color ﬁlm. 42 . near the middle of the range of interesting densities.8/Hm .5. There also seems to be diﬀerence in spectral response of the two meters. if one is happy with Hf being deﬁned in terms of a density of 0. since measuring the luminance of shadows in a scene can be diﬃcult. Therefore. The Minolta Spotmeter F uses a calibration constant of 14. deﬁned by EV = log2 (N 2 /t) = log2 (LS/K). As mentioned in the previous section.16. (52) The ISO standard for color ﬁlm is deﬁned using the same calibration constant. increases exposure and makes a thinner negative. and pp 264268.3 to make the meters match.2 Light meters The calibration problem for color ﬁlm is apparent for light meter calibration. pp 5052. requiring a small value of Hm to produce a 0.1 over background and fog.1 density. expressed as N2 LS = .5 for the current ISO standard. will have a higher speed than will a slower ﬁlm which needs a larger Hm to produce a 0. It follows then that one can use S to determine Hf . Thus a fast ﬁlm. 11. use the current calibration constant of 12. t K which may seem more familiar in terms of the exposure value EV . The upshot is that L/Hf becomes LS/K. such as the newer Sekonic’s. which some ﬁnd more appropriate for color ﬁlm. calibration constants determine the correspondence between densities and exposure. is inconvenient. and on occasion have been know to claim that the Sekonic’s are miscalibrated. and since the range of color ﬁlm is closer to 5 stops. This. Increasing the calibration constant. hence one needs to choose something like Hf = rHm . Newer meters. however.5 to K = 14. however. The Sekonic manuals state this explicity.5. will decrease the EV by 0. where K is a called the calibration constant and is used to encompass both r and the physical constant of proportionality mentioned above. The details may be found in Jacobson(1988). For black and white ﬁlm. N and t and the scene luminance L and ﬁlm speed S. In summary: one has a relationship between the camera controls. There is clearly a diﬀerence between the two calibrations over and above the calibration constant. the range of usable densities is taken to be 7 stops. I ﬁnd it necessary to decrease the EV by about 0. but since color ﬁlm involves three ﬁlms. where r is approximately 3. Older meters use other values. then the ASA ﬁlm speed is deﬁned as S = 0. Hf is usually deﬁned in terms of a larger density. If L is measured in cd/m2 then K = 12. the compromise is not always a happy one. Changing from K = 12. The formula K = LS/2EV may be used with luminance L in cd/m2 and S the ASA ﬁlm speed. but it is easy to obtain for any meter that publishes a table converting EV into luminance.
For N = 16 this gives L=3200 cd/m2 which is the proper value for a midtone in bright sunlight. For larger ranges this will not help. For B&W ﬁlm.3 18% gray cards Since log2 (100/18) ≈ 2. texture but no detail Very dark gray Dark gray Middle gray Light gray Very light gray Near white. Interms of DIN. For a range smaller than 7 EV. texture but no detail Pure paper white Table 4: Zone deﬁnitions 43 . it follows that in any scene an 18% gray card will be placed 2. 11. For a 7 stop range.5 stops down from the maximum brightness. For longer ranges.11. virtually indistinguishable from 0 Near black. 0 I II III IV V VI VII VIII IX Pure paper black Black. where S is the ASA speed. 11. The deﬁniton of exposure value above. The object is to expose and process ﬁlm so that the luminance range of the subject is capatured within this seven zone range. then the middle of the range will be 18% gray. one has N 2 /t = SL/K.5 Contrast control — the Zone system The standard assumes an exposure range of twice 3. Adams (1981) assigns 10 zones as shown in Table (4). If the luminances in a scene span this or a smaller range. or about 7 EV. one may change the characteristic curve by changing processing times. one has t = 10(DIN −1)/10 . If the scene has a range of 5 stops. 18% gray will be placed 4.5 stops up from Hm and therefore an 18% gray will not be in the middle of a 7 stop range.6 EV.5. one may adjust the standard by decreasing the EV. and a lightmeter will place the lowest tone one stop up from Hm .4 Sunny 16 rule The sunny 16 rule states that the correct exposure for bright sunlight should be f/16 at 1/S seconds. and if one takes t = 1/S. then all will fall on a usable portion of the characteristic curve. the lower portions of the range will not be captured by the ﬁlm and both extremes may be lost when printing: the diﬃculty is less serious for B&W ﬁlm with its greater latitude. then one has N 2 = L/K. The Zone System is a methodology for controling the contrast of B&W negatives. the seven zones II through VIII are the only ones of interest. and this may be used to adjust the range to the range of the paper. In practice.
If ∆E is the range of EV’s in the scene to be mapped into a range ∆Z of zones. In addition.6 to obtain equaion (54). Experiental eﬀort is required to ﬁnd suitable processing parameters to achieve G. resulting in overexposure for developing times that compress the range. Davis apparently increments from zone I1/2. as a basis. The argument leading to equation (53) is that since the light meter calibration adjusts EV’s by 3. one must then add (loZone − 1. Since each zone corresponds to 1/r EV units. If ASA7 is the ASA speed when SBR = 7. which changes according to the SBR.6.4 − 1/r. The lighter tones will appear to be blocked using Davis’s formula. DM is the maximum density. EV = (loEV ) − (loZone − 1/2)/r + 4. R is the exposure range. I is the intercept. one must allow for the eﬀect of speed.9. and uses 4 instead of 3. I + R] is 44 .5)/r + 3.6 11.5)/r to obtain the EV that will produce a tone corresponding to lo Zone. and the average gradient over [I. where r = ∆Z/∆E. The appropriate exposure is obtained from EV = (loEV ) − (loZone − 1. x is exposure (log H). He multiplies the average gradient. The assumption being that 7 is a normal range. one can remove this calibration to obtain a base level corresponding to an exposure which will produce a just noticeable tone: presumably this corresponds to zone I+1/2. which he designates by SBR. 11. which is negative for r < 1. (55) where D is density. and F is base plus fog. The gradient at x is D (x) = Dm π sin R pi x − I 2 R cos pi x − I 2 R . then he deﬁnes SBR = 7/r.Davis (1999) describes a systemaic methodology for adjusting the contrast for B&W. G by r to obtain the average gradient required of the characteristic curve in order to place the zones in an acceptable density range for the negative.6 + log2 (r). (54) The diﬀerence between equation (53) and (54) is log2 (r) − 0.6 to produce mid gray tones on a seven stop scale.1 Characteristic curves Graduating function The following function may be used to describe a characteristic curve: D(x) = F + DM sin πx−I 2 R 2 . He uses the subject luminance range. then the ASA speed should be adjusted to r × ASA7 . (53) where loEV is the EV to be placed on the zone loZone.
8 0.56 5.8 5. and than greater exposures will pro¯ duce a decreasing density solarizaiton. I R (56) which diﬀers slightly from that used by Davis (1999).176 4.86 0.7 Test 3 0. ¯ deﬁned by Davis is DR/(Gc log(2)). and Gc is the average gradient obtained from the center range of densities from 0.1+F to DR+0.1+F using equation (55).6 0.796 2. Table (5) gives the ﬁtted parameters plus other derived estimates. The subject luminance range.056 2.9 Test5 0. which in the case of Test 1 is less than DR+0.43 0. It is important to note that the maximum density predicted by the graduating function is M. it suggests that the development process will not achieve higher densities.531 3.2 Data Test 1 0.2 0.77 2.31 9.934 3. Similar plots with other ﬁlms show the same good agreement. The ﬁt is essentially a perfect ﬁt.06 2.1+F.55 2. Film speed is obtained as the log H value corresponding to a density of 0. Figure (23) shows a plot of the ﬁtted values.8 F DM I R Speed Davis’s ¯ G ¯ Gc Davis’s SBR Davis’s Speed ¯ Gc SBR Table 5: Fitted parameters The data in table (6) was copied from page 47 of Davis (1999). SBR.220 3.131 2. Since the graduating function maximizes at this point.88 3. The SBR is therefore taken as M/Gc 45 . The ¯ ¯ G is calculated from equation (56).0. suggesting that the graduating function captures the essence of the underlying mechanism.DM ¯ sin G= R πx−I 2 R 2 DM I +R = .27 0.792 3.1+F to M.8 10.57 0.9 3.805 2. and ﬁtted with equation (55).077 0. In ¯ this case Gc is calculated using the density range from 0. 11.21 0.506 2.069 1.6. where DR is the useable density range which is approximately equal to the paper’s SI: here the DR is taken as 1.72 0.1+F. where only the center range of densities are used.
since 2/s ≈ 2m/f . Thus twice m(h/f ) means either that the object is twice the size m(2h/f ) or twice as close 2m(h/f ). AOV. h. enabling pictures in situations to which the eye is unaccustomed. and it is interesting to discuss it as if it were a camera. Such a camera would have a focal length of about 1 and the image projected on the retina would also be 1 . and m is magniﬁcation. and nearby objects which violate proportionality. It follows that the angle of view.Figure 23: Fitted values 12 Perspective and normal lenses Without the ability to assess distance using binocular vision. where f is focal length. is 2 arctan(1/2) = 53◦ . but it does obey the laws of optics. s. Since m is small. which seems to be in the neighborhood of 10 for normal eyes. The magniﬁcation is thus m = 1/Dn . can be brought into focus. one has arctan h/s ≈ h/s and thus the proportionality also applies to angles. and an object ﬁlling the retina will be of size Dn . 46 . The ratio of object height. the visual system must rely on other clues: the principal one being angular size. The human eye is not a camera. For small angles. to distance. the proportionality holds for the entire practical range of the visual system: for all objects from the nearest focus to the most distant. is h/s = (h/f )[m/(1 + m)] which is approximately m(h/f ) for small m. The assumption that the angular size of an object is proportional to distance is not a bad one. Cameras are not limited to small m. Let Dn be the nearest distance of clear vision.
27 0.41 0.68 0.40 1.73 1.26 0. A larger focal length does the opposite.12 0.log H 3.59 0.39 2.37 1.23 1.09 0.87 0.13 0.10 0.68 0.53 Table 6: Observations and ﬁtted values To appear normal.09 0.74 0.09 0.23 0.55 0. α.16 0.46 0.75 0.76 Test 3 D 0.48 2.59 0.45 0.08 0.11 0.65 0.27 1.15 0.57 1.33 0.9 2. 47 .11 0.90 1.14 0.1 0. A lens with a smaller than normal focal length will have an AOV greater than 53◦ .7 1.73 1.64 0. will appear to be to be X53/AL units apart.92 1.25 1.08 0.77 0.54 1.5 0.73 0.36 1.27 0.03 2.23 0.97 1.77 0. making distant objects appear closer: the consequence of this is that two objects which to the naked eye appear X units apart.97 1. where AL is the AOV for the longer lens.40 0.41 1.18 1.61 Test5 D .54 1.01 3.45 2.55 0.11 0. then an object occupying the proportion α will also have angle 53α◦ .09 0.19 0.45 0.58 0.36 0. For shorter lenses with AS .60 0.33 0.19 0.38 2.45 1. of course.08 1. still appear reasonable to the eye because binocular vision provides triangulation information to the visual system.64 0.89 2.17 1.59 0.06 1.16 0.95 0.17 0.08 0.20 0.11 0.3 2.44 0.05 2.67 0.88 2.75 2. An image occupying a proportion.15 2.07 1.10 0.50 0.25 1.08 0.25 0.07 0.40 0.09 0. a picture held at Dn should produce the same angles for objects as does the image that forms on the retina.77 Fit 1 0.49 0.32 0.55 1.50 0.69 0. of the picture will have an angle of 53α◦ .73 0.85 1.58 0.15 2 1.58 1.2 0. even when they cannot be brought into focus.19 0. It should be noted.07 0.09 0.62 Fit 3 0. so the objects appear to be XAS /53 units apart: the separation appears greater than the naked eye expects.15 0.4 1. In addition.34 0.35 0. producing even greater exaggeration.17 2. the AOV.31 0.55 Fit 3 0.32 0.29 2.27 1. short lenses are more frequently used for nearby objects for which proportionality does not hold.36 0. If a camera lens has this AOV. and thus appear normal on a print. that such nearby objects.6 2.25 0.27 2.50 0. making an object occupy a smaller angle and thus appearing farther away. the situation is reversed.87 0.41 0. It follows then that normal lenses are those whose focal lengths are approximately equal to the diagonal of the ﬁlm size.46 1.47 2.8 0.07 1.12 0.08 0.05 Test1 D 0.23 0.
corresponds to 127 l/in.13 13. but a slight degradation can be detected. A line on these charts is the extent of a black and white bar. tests using photographs of a USAF test target with the Epson Photo EX shows that this printer requires a value between 200 and 240 dpi. must be determined from the print requirements. or about R/75 l/mm. this means that about 250 dpi must be sent to the printer. How much is this? Obviously it varies with the purpose. it need not be more than the human eye can use. and thus the captured rf becomes rf = R/75 = (T /df )/75 = (225dp /df )/75 = 3M in l/mm. From this. (1) A careful examination of scans a USAF test target. the captured ﬁlm resolution rf is about R/3 l/in. let us choose 225 dpi as the required dot density for a printer. The minimum distance of distinct vision is about 25 cm. R. and the 200 dpi is almost as good. It follows that for a scanning density of R dpi. (2) Digital to print. If one sends 300 dpi to the printer. then rf = 4M . where M = dp /df . the captured rf is shown in Table 7 for several ﬁlm and print sizes. while with 3 there will be at least one on each color. What is the corresponding resolution on the ﬁlm? There are two parts to the problem: (1)Scan to digital. the resolution of the human eye. (2) The density. 4x5 12 6 3 8x10 24 12 6 16x20 51* 24 12 R 3000 2000 1000 R/75 40 27 13 35 6x7 4x5 Table 7: Captured ﬁlm resolutions in l/mm given print size * Not possible with available scanning rate 48 . which is the distance between the frontiers of one line and its successor. Taking rf = 3M . suggests about 3 dots seems to be the minimum number required to separate lines. but printers need 2 input dots to show a USAF test target line. as one would to with a LightJet 5000. In bright light the human eye at this distance can just resolve 5 l/mm which sets a limit on the print resolution required. but for photographs. Appropriate scan rates for each ﬁlm size and the maximum onﬁlm captured rf is also shown. where dp and df are corresponding dimensions of the print and the ﬁlm. For convenience. one has R = T /df . The total number of dots required by a printer is T = 225 × dp = R × df . If one chooses 2 dots instead of 3 in paragraph (1). The 240 dpi is excellent. then rf = 4M or 5M . the line may be missed because both dots fall in the black or in the white. Large commercial printers often require 300 dpi. With two dots.1 Digital Scanning and resolution The aim of scanning is to extract that amount information from the source material required by the output material. Now 5 l/mm. As conﬁrmation. Which means that most people cannot focus on closer objects.
2 Digital cameras The arrays in digital cameras are much smaller than ﬁlm sizes. In the scanned image. so 9 × 5 = 45. then the resolution from a 3000 dpi scan is 3000/150 = 20 l/mm. which seems reasonably consistent with these observations. and the ones on the leaves perhaps 8 l/mm. The names 1/3 inch. the resolutions for all formats represent the same degree of detail in the subject: e. since the major ﬁne lines at 45 l/mm are clear with a 20 power loupe. The commonly used ones are shown in Table 8. Therefore given a ﬁxed print size. Figure 17 on page 33. are a holdover from previous technology and do not represent the actual size of the arrays. The loss can often be compensated by sharpening the edges with unsharp masking.g. A 150 mm lens was used at a distance of 1500 mm. since the resolution is low in this printing. This is not true for a 16x20 print. etc. The resolution of the ﬁlm is near 60 or so l/mm. The diﬀerence is traceable to the fact that 3 dots are needed by the scanner for a digital print. for example. the major ﬁne lines can be seen against a black background. but the lines in the leaves at about 72 l/mm are not distinguishable12 . The dimensions have been rounded to the nearest millimeter. where there is a diﬀerence between 35 mm and 4x5 because the 35 mm rf is limited to 40 l/mm (= 3000/75) instead of 48 l/mm.One sees. that the minimum captured detail in an 8x10 digital print from a 4x5 transparency is about 6 l/mm. This is less than the minimum resolution for a conventional print of about 10 l/mm. It is useful to note that four times the 4x5 rf is equal to the 35 mm rf . height 4 5 6 18 width 5 6 9 28 pixels 480x640 1000x1300 1400x1700 2000x3000 1/3 inch 1/2 inch 2/3 inch large Table 8: Approximate digital camera array sizes in mm 12 Comparing the line width with that of an Edmund chart shows the major lines are about 5 l/mm. shows a stack of dollars which were photographed and scanned at 3000 dpi — it is printed here only to show the test setup. The photo used was 1322 in my ﬁles. nor can the lines in the leaves be separated. If one requires 3 dots instead of 2. 49 . but they cannot be separated when near each other as in the wreaths. 13. I conclude that one needs more than two dots to make a line clearly distinguishable. As a caution. an object just resolved in an 8x10 print from a 35 mm slide will be just resolved in an 8x5 print from a 4x5. giving a multiplying factor of 9. Scanned photographs of the USAF chart conﬁrm that 3 dots are needed at the highest detectable frequency. there is always much more resolution in a ﬁlm than may be captured in a scan.
. EV . scale.4. A rational scale of values corresponding to aperture diameters13 is the exposure value. in which case a small opening and large fnumber represents a bright object. and markings on the lenses were likely to have been in terms of fnumbers. There is not a lot that can be said in justiﬁcation of this standard sequence for a practical photographer. 4. 50 . as early photographers quickly learned.1 Miscellany Fractional stops The amount of light admitted through the aperture is of great importance. 2. From equation (52). Tests on the Olympus C2500L with a 2/3 inch array produced 60 to 68 l/mm depending on the focal length used. one can see the relationship between EV and the fnumber. since the opening is a polygon. but the odd sequence is still something to remember. It may be seen that only the large array is adequate for 8x10 prints.1994). The EV value 2. One way to think of it is as a measure of the luminance of the object. It may be controlled by the duration of the exposure and by the lens aperture. or EV = 2log2 (N ). A limited test seems to conﬁrm the accuracty of this table. The R in this table is the number of pixels per inch for the shorter dimension of the array. and it increases as the light admitted decreases.} is fairly recent. which is contrary to what one wants. Iris shutters were also introduced about 1880 (ibid).8. not a circle. Sequences of stops and fnumbers were proposed at the turn of the century. 2. It is this 13 The blades of iris diaphragms are curved to produce an aperture scale with equidistant spacings.8. 14 14. (Ray. About 1880 according to (Kingslake. N by setting the time t to unity: it is N 2 = 2EV . The movement may not be completely uniform for intermediate stops.1/3 inch 1/2 inch 2/3 inch large 3x5 57* 45 38 13 5x7 95* 76* 63 21 8x10 152* 122* 102* 34 R 3200 5000 6000 2820 R/75 43 67 80 38 Table 9: Captured digital array resolutions in l/mm given print size * Not possible with available array dimensions A table of captured resolutions for these cameras is shown in Table 9. .5 corresponds to a point between f /2 and f /2. It requires an act of memory to use. 1989) it became customary to designate a lens by its fnumber. This agrees with the captured resolution in Table 9 of 63 l/mm. . but the present standard sequence of fnumbers {1. 1. In general fractional stops on large format lenses should not be relied on to less than 1/3 stop and probably not to 1/2 stop.
then U = V /(1+m) is the fnumber allowing for the bellows extension. 51 .4 5. Cameras with automatic diaphragms can be expected to adjust their apertures to this scale.4 + 1/2. one may desire the number of stops associated with the bellows factor. and 1. 15 Remember that a decrease in EV corresponds to a decrease in fnumber. Thus 2EV +0. or the tables in the Sekonic Zoom Master L508. where 1/T is the shutter time in seconds. This is given by mr .1 step (Ray. one ﬁnds that the shutter speed necessary to stop this motion is T = rf r =m . Useful notations for the printed page are 1. thus the increase is %41 not %50. c(u − f ) c where m is magniﬁcation and r is in millimeters per second. 14. m. c(1000u − f ) It may also be useful to cite the distance the image of a point moves during 1/T seconds.04rf .4”. The amount of light actually admitted through the diaphragm is not. is not zero. and an increase in aperture size.0.5 than at EV = 2.g. or L608.2 Stopping Motion Let c be the diameter of the circle of confusion.5 = (LS/K) 2 ≈ (LS/K)1. that is 2EV1 = V 2 .41.45 with the “5” in smaller type than the “1.“fractional” value that appears on digital light meters14 : e. From equation (52). s u−f and if an object moves at a rate r then s = r/T . cT where the distance is in units of c. at least to a 0. however. Thus one opens up log2 (1 + m) stops15 from V to obtain the correct fnumber U . and f and c in millimeters. and s the corresponding subject size at distance u for a lens with focal length f . one has √ 2EV = LS/K. u in meters. Suppose EV1 is the exposure value corresponding to V . then U 2 = 2EV1 /(1 + m) = 2EV1 −log2 (1+m) . then T = 447. When magniﬁcation. then c f = . 1994). If V is the fnumber when focused at inﬁnity. 14 See the Minolta Spotmeter F instruction manual. %50 more at EV = 2. If r is in mph.5 = (LS/K)20. a meter reading might be 1.
14. The camera must be moved to accommodate this. 4 If e is the extension of the bellows.4 Stops as distances The inverse square law gives the ratio (d2 /d1 )2 for light intensity at two distances d1 and d2 . is the angle subtended at the center of the lens by either the object ﬁeld or the image extent. Thus a bellows with extension equal to the focal length of the lens delivers 1 the light. 14. 14. and the number of stops is the base 2 log of this ratio. 2 tan(A/2) 52 . or exposure. d . or the fnumber can be divided by BF . as one with zero extension. since v = f (1 + m) = changes with d. where m is magniﬁcation.5 AOV The angle of view. where d is the image extent. e = 250 with f = 200 produces the 45 and 20 fnumbers. which is a bit more than 2 stops. then magniﬁcation is unaﬀected for a constant AOV. 2f tan(A/2) which shows that if f is scaled by the same scaling as d when format is changed. It is useful to note that this may be rearranged as. and f the focal length. then the bellows factor is deﬁned as BF = f +e f 2 = (1 + m)2 .3 stops given by log2 (BF ). This may be compared with the actual value of 2.3 Bellows Factor The inverse square law applies to bellows movements: the intensity of the light falling on a surface is inversely proportional to the square of its distance from the lens. Simple trigonometry shows: A = 2 tan−1 d 2f (1 + m) . 1+m= d . m the magniﬁcation. hence 2log2 (d2 /d1 ) is the change in stops. The time value (1/sec) is multiplied √ BF to adjust for the decrease in by light. A. An easy procedure is to divide f + e and f by something that brings them into the conventional fnumber range and then to count the stop diﬀerences: thus.
the image will be of size 2c. 14. Equation (34) may be used with c = 1/L: L= 2N (1 + m) . A test must be run to justify 1/f in the ﬁrst place. δ where δ is the depth of focus along the rail. and L is lines per millimeter.7 Hand holding a camera Taking 1/f as the exposure time is a common rule for hand holding. The diﬀerencial of δ with respect to L becomes δ dδ = − dL = −δc dL. Suppose that vibrations from a lens of focal length f . since the ﬁlm backing is presumably ﬁxed. This suggests that ﬁlm movement disturbs the image at the exact focus more than at the DOF limits.9 Miscellaneous equations Magniﬁcation as a function of nearDOF .6 Film ﬂatness and vibration eﬀects The eﬀect of a lack of ﬁlm ﬂatness or vibration may be assessed in terms of changes in the depth of focus. 14. can be damped by an exposure time of 1/f . L which is the amount of ﬁlm movement which degrades the image by one line per millimeter — the smaller the dδ the more sensitive. c. It is likely that a parameter will be needed to reﬂect individual variation. Fix the magniﬁcation so that the subject will always image the same size on the ﬁlm. ∆ (57) . This means that a point in the subject will image as a circle of confusion. m= N c(f − ∆) + f∆ 53 N c(f − ∆) f∆ 2 + Nc .Hence one should cut the vibrations in half by using 1/(2f ) as the time of exposure to obtain an image of size c. one sees that the 1:1 focus point is attained when u + v is a minimum. 14. It follows that a 2f lens puts the image twice as far from the lens.14. One should probably use half of this. In practice one could use a measurement from a mark on the rear standard to a mark on something that moves with the subject and vary the subject distance until the minimum infocus distance is found.8 Finding the 1:1 focus point 2 u Since u + v = u−f which has a minimum at u = 2f . and with the same amount of vibration. It has some justiﬁcation: First assume that halving the time of exposure will halve the size of the blur.
Focal length as a function of DOF. Nc m DOF . f= where M = (m + 1)/m2 .where ∆ = u − Zn is nearDOF . DOF − 2N cM (58) 54 .
Ray.. Harold.G. (1994) Applied photographic optics. (1999). Focal Press. Academic Press.E. Kingslake. (1979) Lens design fundamentals. NY. (1988).BIBLIOGRAPHY 1. Leslie. Richard. (1996) View camera focusing in practice. Reading MA. Addison Wesley. 55 . Sears. 11. S. Focal Press. 8. Canada. Boston. Hansma. (1993) Focusing the view camera. Photo Techniques. Adams. R. John. Mouroulis. 7. 4. Focal Press. and Zakia. Focal Press. and Attridge. 9. Merklinger. Paul K. 5. Kingslake. Ottawa. (1958) Optics. Beyond the Zone System. and Macdonald. Compton. Phil. Boston. S.F. Boston. 10. Stroebel. Current. Oxford. Boston. 6. G. 2. Jackobson. (1997) Geometrical optics and optical design. Ira. R. 172. Academic Press. F. Oxford 3.F. (1981) The Negative. Boston. 5457. The manual of photography. (1986) Photographic materials and processes. (1989) A history of the photographic lens. Little Brown. P. Ansel. M.. J. Ray. Oxford. R.W. Davis.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.