►
Description
We will have a talk this month from John Hayes. Here's the talk description followed by John's Bio. There will be Q&A during/after the talk.
Optics Adventures During the Pandemic: Engineering a Remote Imaging Telescope
This is the story of engineering a remote 0.5 m telescope that will be deployed in Chile for astronomical imaging. Getting this system configured, tested, and running turned into a year-long project while in pandemic isolation. This project drew on nearly every optical engineering discipline including optical design, physical optics, optical testing, mechanical design, sensor optimization and systems engineering—along with some “garage-shop” problem solving. A new AI based wavefront analysis system was also used to precisely evaluate optical alignment. We’ll start with a review of the optical design of a 20” Planewave CDK telescope, look at the optical system for guiding and auto-focusing, and review some of the considerations for selecting an appropriate imaging sensor. Then we’ll look at what it takes to actually configure, test, align, and operate the telescope as a part of the commissioning process in the real world. Along the way, we’ll encounter a number of practical lessons that apply to any optical engineering project--whether amateur or professional.
Dr. John Hayes received his BS in Physics and Astronomy from the University of Arizona, in 1975. He received MS (1981) and Ph.D. (1984) degrees in Optics from the Optical Sciences Center at the University of Arizona. Dr. Hayes became involved with WYKO Corporation shortly after its inception in 1982 and spent 15 years as a member of the technical staff, senior management team, and corporate board. As a Research Professor at the College of Optical Sciences (1999-2001,) Dr. Hayes developed interferometric vibration-compensated metrology solutions for aerospace and large telescope applications. In 2001, Dr. Hayes cofounded 4D Technology Corporation where he served as president through 2006 and a board member through 2018. The author of many technical publications, Dr. Hayes also holds numerous patents in the areas of optical testing, atomic force microscopy, suspension arm metrology, and dynamic interferometry. Many of these technologies have been used to fabricate optics for the largest telescopes in the world including the James Webb Space Telescope. In his spare time, he operates his own remotely controlled 0.35 m telescope in New Mexico for imaging deep sky objects. He is currently configuring a new 0.5 m system that will operate remotely under the southern skies in Chile. He has been recognized with numerous IOTD awards and a NASA APOD for his images and enjoys using his engineering background to further explore and advance the state of the art in amateur astronomical imaging. He currently serves as chairman of the James C. Wyant College of Optical Sciences advisory board.
A
So
hi
everyone
welcome
to
the
august
meeting
of
the
sjaa
obser
imagers,
and
tonight
we
have
john
hayes
going
to
talk
to
us
about
a
remote
telescope
in
chile.
John's
been
a
optical
professional.
Should
I
say
is
that
the
right
word
for
his
whole
career,
an
expert
in
optics
and
has
turned
into
an
astrophotographer
reverted
from
his
original
film
days
to
a
or
a
photographer
and
as
an
expert
on
and
all
sorts
of
stuff
that
we're
going
to
hear
about.
B
Okay,
well,
thank
you,
hi,
I'm
going
to
go
ahead
and
we'll
see
if
we
can
get
the
screen
sharing
going
here
and
see.
If
we
can,
can
you
guys
see
that
there?
Yes,
there
we
go
okay,
we're
good!
Well,
I
want
to
thank
you
guys
for
inviting
me
to
do
this.
You
know
I
love
talking
about
imaging
telescopes
and
as
an
optical
engineer,
you
know
this
is
there's
nothing
more
fun
to
me
than
building
telescopes
and
for
me
you
know
a
lot
of
people
image
and
and
enjoy
it.
I
was.
B
B
So
this
is
a
the
story
of
this
20-inch
telescope
that
I
configured
I
bought
and
configured
during
the
pandemic,
and
I
will
tell
you
the
whole
story
here:
let's
see
if
we
can
oops,
let's
say
just
take
me
a
second
here,
laser
pointer,
I'm
going
to
get
a
laser
pointer
going
and
there
we
go,
and
I
should
be.
Oh
there
we
go
so,
let's
just
a
little
quick
overview
of
what
we're
going
to
talk
about
is
you
know
the
motivation.
B
B
What
I'm
going
to
do
for
guiding
and
focusing
and
then
how
I
went
about
evaluating
the
performance
of
this
thing,
and
I
did
some
interesting
things,
some
of
them
old,
older
technologies,
some
of
them
newer
technology
and
we'll
go
over
flat
correction
image,
calibration,
vignetting
and
baffling
how
it
worked
in
this
and
then
sort
of
the
problem.
Solving
part
of
this.
You
know
nothing
often
in
engineering
projects.
It
doesn't
always
go
the
way
you
planned
and
you
reach
a
point
where
you
have
to
solve
problems,
and
that
was
no
exception
here.
B
I
haven't
had
an
eyepiece
on
a
telescope
for
a
long
time,
and
so
I
am,
I
am
purely
an
imager
and
I'm
glad
to
hear
that
most
of
you
guys
are
as
well.
So
you
know
that
that
this
is
a
big
subject,
so
we're
gonna
go
over
a
few
of
the
things,
but
we're
gonna
miss
a
lot
of
it,
but
we'll
talk
about
the
relevant
stuff,
the
stuff
that's
relevant
to
this.
B
This
story-
and
I
wanted
to
start
out
with
the
current
telescope
that
I
image
with,
because
that's
going
to
ultimately
serve
as
our
reference
system.
I
use
a
14
inch
flat
field.
Sct,
it's
a
it's
an
edge
hd
system,
it's
highly
modified.
This
is
not
a
standard
factory
system.
Of
course,
you
guys
know
it's
a
schmidt
castle
grain
that
hasn't
has
a
built-in
field.
Flattener
the
focal
length
is
just
shy
of
about
four
meters
operating
at
f
10.8.
B
As
you
guys
know,
we
get
colors
from
using
a
color
color
filters
in
a
wheel,
filter,
wheel,
changer,
and
you
can
see
that
on
the
back
of
the
telescope
back
here,
I
have
it
sitting
on
an
on
a
astrophysics
1600
mount
which
is
a
pretty
big
mount
for
this
telescope.
But
it
is
absolutely
rock
solid.
You
can
tap
it
and
nothing
moves.
It's
a
very,
very
solid
telescope.
B
It
I
used
to
run
it
outside
of
my
my
hanger
here
in
in
bend
oregon,
and
this
is
the
way
it
used
to
look
on
a
little
roll-out,
pier
that
I
have
it's
not
such
a
little
pier.
This
is
actually
a
massive
pier
and
it's
on
wheels,
and
I
would
just
roll
it
out
and
use
it
here,
and
I
got
it
running
here
and
then
I
moved
it
out
to
dsw
in
in
northern
new
mexico.
B
Just
about
I
don't
know,
maybe
30
40
miles
outside
of
santa
fe,
and
it
works
pretty
well,
but
again,
this
will
serve
as
our
reference
system
for
this
20
inch
scope,
see
if
I
can
get
the
system
kind
of
hangs
up
here,
hang
on
with
me
for
a
second
there
we
go
for
those
of
you
who
haven't
done
remote
imaging.
I
stuck
this
slide
in
here,
so
you
can
see
what
a
remote
imaging
observatory
looks
like.
This
is
just
from
one
of
the
the
observer,
the
cameras
in
the
observatory,
my
scope.
B
B
They
of
course,
they
provide
power,
network,
weather
reporting,
video
monitoring
and
roof
control.
Their
customers
include,
you
know.
Private
individuals,
like
myself,
government
and
research,
research
organizations
and,
in
fact,
there's
a
scope
here.
I
think
it's
this
one
over
here,
there's
a
celestron,
that's
used
by
folks
who
are
monitoring
chinese
missile
launches
and
they're.
Looking
for
orb
they're,
trying
to
calculate
orbital
parameters
with
this
particular
scope
and
they
run
it
with
a
a
hyperstar
system
and
it's
on
every
night.
B
So
a
big
question
is
you
know
why
get
a
bigger
scope?
Well,
the
advantages
are
potentially
more
signal:
better
signal-to-noise
ratio
potentially
more
image
detail
and
that
appeals
to
me
a
lot
and
potential
for
a
larger
field
of
view.
These
some
of
these
larger
scopes
are,
I
have
a
faster
focal
ratio,
shorter
focal
length
and,
with
my
sensor,
could
potentially
give
me
a
little
bit
bigger,
bigger
field.
Of
course,
the
downside
is
all
you
guys
know
these
things
cost
a
lot
of
money.
B
You
know
the
20
inch
plane
wave
scopes
are
about
almost
five
times
more
expensive
than
a
14
inch
celestron
and
the
24
inchers
are,
you
know
almost
nine
times
a
little
more
nine
times
as
much
as
a
14
inch.
I
mean
they
get
pretty
expensive
and
of
course,
the
biggest
thing
is
they're
more
affected
by
atmospheric
turbulence,
and
probably
one
of
my
biggest
kind
of
issues
that
I
deal
with
out
at
dsw
is
the
seeing
and
the
seeing
out
there
the
median
thing
out
there-
and
I
don't
have
a
study
to
show
this.
B
But
my
sense
is
the
meetings
median
seeing
is
around
two
arc
seconds.
You
know
they
a
lot
of
these
observatories,
love
to
tell
you
the
best
they've
ever
had
and
you'll
hear
numbers
like
one
arc
second
out
there,
and
I
have
seen
one
arc
second
out
there,
but
it's
extraordinarily
rare.
So
the
site
selection
is
a
big
deal
and
that
put
me
off
of
buying
a
big
telescope,
a
bigger
telescope,
rather
for
a
long
time.
B
Well,
I
was
at
a
meeting
a
few
years
ago
and
these
folks
from
ops
tech
showed
up.
They
have
a
they're,
a
remote
telescope
hosting
facility
in
chile,
about
20
miles
south
of
where
the
gmt
is
located
in
the
atacama
desert
and
they
showed
up
and
they
said
gee.
We
have.
You
know
320
over
320
nights
of
clear,
clear
nights
a
year
I
put
down
300
nights
here
on
this
slide.
That's
still
plenty
good
enough
for
me
and
they
claim
the
median
seeing
is
about
a
half
arc.
Second,
which
is
really
world
class.
B
I've
heard
reports
it
can
reach
0.1
to
even
0.2
arc
seconds
of
seeing
that
just
seems
unbelievable
to
me,
but
I
have
heard
credible
results
that
they
actually
can
reach
that,
and
I've
heard
from
users
out
there
that
a
half
arc
second
certainly
below
an
arc.
Second,
is
very
common,
so
my
initial
goal
with
this
telescope
was
to
make
sure
that
it
could
whatever.
B
However,
I
configured
it,
that
it
should
permit
at
least
a
half
arc
second
performance,
and
ultimately
I
wanted
to
buy
something
that
at
least
doubled
the
collecting
area
of
the
telescope,
which
meant
you
know
a
20-inch.
You
know
making
it
jump
to
a
17-inch
or
a
16-inch
from
the
14-inch.
Wasn't
a
big
enough
jump
for
me.
The
20-inch
was
the
next
size
up
that
I
would
even
consider-
and
ultimately
I
decided
on
the
20
inch
and
mainly
a
big
part
of
that
was
the
cost,
but
the
other
part
of
it
was
you
know.
B
Not
only
does
the
cost
of
the
bigger
telescope
get
more
expensive
that
the
cost
of
everything
that
goes
with
it
and
the
difficulty
of
everything
goes
goes
with.
It
goes
up.
I
mean
if
I
gone
to
a
24,
you
know
the
weight,
the
size.
The
complexity
of
the
whole
system
really
takes
a
big
jump
and
it's
totally
doable,
but
even
with
this
you
know
I
had
to
had
to
change
a
lot
the
way
I
do
a
lot
of
things
with
this
big
telescope.
This
is
the
plane
wave
system.
B
B
They
claim
a
well-corrected
field
over
a
52
millimeter
image
circle,
which
is
just
fine
for
me
with
me.
They
have
few
silica
mirrors
with
low
thermal
expansion.
It's
all
carbon
fiber
and
you
know
they
claim
it's
easy
to
align
and
yeah.
I
think
it's
easier
than
an
rc
to
align.
That's
for
sure.
Central
obscuration
from
this
large
secondary
is
pretty
big.
It's
almost
40
percent,
so
that's
pretty
big
and
high,
and
I've
talked
about
back
working
distance
being
a
really
important
thing.
This
says
5.8
inches,
but
that's
with
their
focus
here.
B
It's
actually
8.1
inches
and
that's
a
critical
number
and
it
takes
a
little
bit
of
digging
to
get
the
right
number
out
of
those
guys,
but
they
they
do
put
a
number
up
here
and
I've
I've
explained
to
them.
They
need
to
put
that
number
up
for
all
of
their
telescopes
and
I
think
they
heard
me
so
hopefully
that's
going
to
become
a
regular
thing.
B
One
of
the
cool
things
about
the
20
inch
is
that
the
design
is
available
in
in
a
book
and
the
telescopes,
astrographs
and
eyepieces
is
a
is
a
great
book.
It's
now
out
of
print,
unfortunately,
but
it
goes
over
lots
of
different
telescope
designs
and
the
plane
wave
20
happens
to
be
in
there.
So
I
could
stick
it
into
into
oslo
and
ray
trace
it,
and
this
is
what
it
looks
like.
It's
got
an
f3
primary
with
2.27
magnification
spherical
secondary
you'll
notice.
It
does
have
some
vignetting.
B
These
are
actually
the
size
of
the
field,
correcting
lenses,
and
so
there
is
some
v,
some
vignetting
in
the
system,
but
it
produces
really
nice
spot
diagrams,
and
I
you
know
we
could
look
at
wavefront
and
all
that
stuff,
but
I
don't
want
to
go
into
all
of
the
the
the
components
of
understanding
this
design
in
this
particular
talk.
But
you
can
see
these.
The
black
circle
is
the
size
of
the
arie
disk
in
in
this
display
of
spot
of
geometric
spot
diagrams,
and
so
on
axis.
B
It's
clearly
diffraction
limited
even
out
to
point
seven
seven.
Seventy
percent
of
the
field
which
which
goes
to
the
edges
of
a
52
of
a
36
by
36
millimeter
sensor,
it's
still
pretty
much
diffraction
limited.
It
starts
to
fall
apart
a
little
bit
in
the
corners
of
a
52
millimeter
circle,
but
still
very,
very
good
and
and
with
respect
to
any
atmosphere.
It's
seeing
this
will
produce
pinpoint
stars
across
that
52
millimeter
field,
image,
circle
and,
of
course,
a
little.
B
This
image
came
to
me
from
some
guy
on.
I
can't
remember.
I
think
it
was
on
astrobin,
saying
gee.
I've
got
something
wrong
with
my
stars.
What
is
this?
Well,
of
course
that's
what
coma
looks
like,
and
if
you
see
this
stars
that
look
like
this
near
the
center
of
your
field,
you've
got
a
misaligned
telescope.
So
it's
a
good
idea
to
be
able
to
have
some
idea
of
how
to
how
to
look
at
your
star
images
to
get
some
clue
as
to
how
well
aligned
it
is.
We
all
know
they're
perfectly
round.
B
B
When
we
talk
about
extended
sources
is
the
camera
equation
and
the
camera
k
equation,
falls
out
of
basic
radiometry
and
and
the
basic
conclusion
of
the
camera
equation
is
that
a
faster
system
always
produces
more
irradiance
in
the
focal
plane
and
the
equation
for
the
the
camera
equation
is
pretty
simple
and
I've
just
shown
it
here
as
a
function
of
field
angle.
We
know
that
this
is
just
basically
the
source
radiance,
the
focal
ratio,
the
the
the
out,
the
irradiance
in
the
focal
plane
varies
as
the
inverse
square
of
the
focal
ratio.
B
They
they
tend
to
think
the
image
irradiance
falls
off
only
because
of
vignetting,
but
there's
actually
this
radiometric
falloff
as
well,
and
so
the
main
takeaway
is
the
irradiance
for
extended
objects
is
proportional
to
the
inverse
of
the
focal
ratio
squared
and
what
that
means
is
functionally
an
8
meter.
F4
telescope
produces
the
same
irradiance
in
an
image
as
an
80
millimeter
f4
telescope
for
any
given
extended
object.
You
know
that
that's
kind
of
a
very
interesting
result.
A
lot
of
people,
don't
think
about
that.
The
larger
image,
however,
produces
a
much
bigger
image.
B
The
irradiance
is
the
same,
but
the
image
is
much
much
bigger
in
a
larger
telescope.
So
you
know
that's
kind
of
the
first
thing
that
we
can
see.
You
know
getting
a
bigger
telescope
gives
you
a
bigger
image
for
a
given
focal
ratio
and
but
we're
dealing
with
the
exact
same
irradiance.
The
irradiance
when
we
talk
about
stars
is
a
little
bit
different
and
we
do
the
calculation
slightly
differently
and
I'm
not
going
to
go
through
that.
It's
a
diffraction
calculation,
but
it's
proportional
to
the
diameter
divided
by
the
focal
ratio.
B
This
particular
outcome
is
that
a
bigger
aperture
almost
always
reveals
fainter
stars,
and
so,
and
you
can
see
that
in
a
lot
of
images.
You
know
if
you
look
at
a
telescope
of
the
same
field
with
a
small
telescope
versus
a
large
telescope.
The
large
telescope
almost
always
will
show
lots
and
lots
of
stars
compared
to
the
smaller
telescope,
but
irradiance
in
the
image
plane
isn't
really
what
counts.
B
Noise
varies
as
the
square
root
of
the
responsibility
of
the
responsivity
times
the
number
of
photons
that
are
incident
on
that
sensor.
So
the
signal
varies
linearly
with
the
pixel
dimensions
and
with
respect
to
photon
noise,
the
signal-to-noise
ratio
always
improves
with
bigger
pixels.
That's
the
sort
of
the
conclusion
of
of
all
of
this,
so
in
general,
it's
better
to
have
bigger
pixels
than
smaller,
pixels.
Okay,
so,
and
and
of
course,
when
I
say
this-
I'm
assuming
equal
sensitivity,
okay
and
equal,
read
noise
and
equal
dark
noise.
B
So,
there's
a
little
bit
of
complexity
that
to
that,
but
in
general,
what
we
usually
care
about
in
imaging
is
photon
noise.
That's
the
main
thing
that
we
care
about
now.
I
know
if
we're
doing
lucky
imaging
and
really
short
exposures.
We've
got
to
worry
about
that
read
noise,
but
in
for
most
of
us
doing
long
exposure
low
light
imaging.
B
We
don't
really
need
to
worry
about
the
the
read
noise
or
the
dark
noise
noise
in
terms
of
the
the
most
desirable
determining
the
most
desirable
pixel
size.
So
in
general
you
know
what
works
best.
You
know
if
the
pixels
are
too
small,
then
the
signal
the
noise
suffers
and
if
they're
too
big
the
ms
image
kind
of
quote,
unquote
the
the
image
sharpness
suffers.
Okay,
you
know
if
the
pixels
are
too
big.
B
So,
let's
first
talk
about
how
sampling
works
and
we're
going
to
talk
about
something
called
nyquist
sampling
and
I'm
going
to
kind
of
wave
my
hands
at
this,
because
nyquist
sampling
is
actually
very
easy
to
work
out,
but
it.
But
it
involves
knowing
some
fourier
theory,
and
I
don't
want
to
go
into
all
that
math,
and
so
I'm
going
to
tell
you
just
some
results.
B
It
can
also
do
some
things
in
normal
photography
which
which
relate
to
causing
aliasing
of
high
frequency
information
in
the
image
which
results
in
moire
effects.
If
you
go
out
with
your
camera
and
photograph,
the
side
of
a
building
that
has
a
periodic
structure,
if
you
do
it
just
right,
you'll
see
that
you'll
get
a
moire
pattern
for
periodic
structures.
B
That's
not
such
a
big
problem
for
for
astro
imaging,
but
it
is
one
of
the
effects
of
under
sampling
if
we
over
sample,
if
we,
if
we
just
pick
a
a
camera
that
was
really
really
really
tiny
pixels
so
such
that
we
have
more
than
two
samples
per
cycle
for
the
highest
spatial
frequency
made
available
by
the
optics,
we
can't
recover
any
additional
information.
Okay,
so
remember
this
is
key
and
a
lot
of
folks
who
are
into
electronics
don't
realize
this
an
electronic
system.
We
don't
necessarily
have
a
bandwidth,
limited
system.
B
However,
in
optics
we're
really
lucky,
because
an
optical
system
like
a
telescope
actually
has
a
bandwidth
limit.
Okay,
it
cannot
pass
spatial
frequencies
higher
than
that
bandwidth
limit.
So
it
is
a
bandwidth
limited
system,
so
we
can
easily
tell
where
that
limit
is,
and
if
we
go
beyond
two
samples
per
cycle,
we
don't
get
any
additional
information
out
of
the
system.
B
Critical
sampling
is
the
optimum
sample
sampling
required
for
perfect
image.
Reconstruction
now
remember.
The
theory
requires
point
samples
and,
in
practice
we're
using
little
rectangular
samples.
Okay.
So
it's
not
really
quite
right.
The
theory
isn't
quite
right
for
what
we're
doing,
however,
finite
pixel
sampling
modifies
things
in
a
small
but
predictable
way
and
we're
gonna.
B
We're
gonna
see
that
as
I
talk
about
this
a
little
further,
so
I
want
you
to
remember
that
we're
not
quite
using
exactly
nyquist
but
nyquist
is
a
good
starting
point
to
determine
how
we
should
sample
the
image
and
an
mtf
curve.
Really
what
it
does
is.
It
shows
how
spatial
frequencies-
and
these
are
what
spatial
frequencies
look
like,
they're,
basically
sinusoidal
intensity
patterns.
They
show
it
shows
how,
for
example,
this
high
spatial
frequency
transmits
through
the
optical
system
and
how
its
contrast
is
modified
by
the
optical
system.
B
So
you
know
we
in
in
amateurs
in
in
in
imaging
circles,
we
typically
just
you
know.
We
talk
our
way
around
that,
but
I
just
want
to
emphasize.
This
is
what
an
mtf
curve
an
mtf
looks
like
and
it's
what
the
system
is
doing
to
these
spatial
frequency
content
and
at
the
highest
end.
This
represents
this
point,
represents
the
highest
spatial
frequency
transmitted
by
the
optical
system,
which
is
in
normalized
in
in
actual
space,
rather
in
in
inverse
line
pairs
per
unit.
B
Distance
is
one
over
lambda
times
the
focal
ratio-
and
this
is
also
it
requires
four
point-
eight
eight
point
samples
across
the
arie
disk
in
order
to
get
to
that
point
right
there.
So
if
you
want
to
learn
more
about
this,
this
is
by
the
way
this
is
a
monochromatic
mtf.
This
is
not
a
polychromatic
mtf.
B
If
you
want
to
learn
a
little
bit
more
about
this,
I
did
a
complete
talk
on
the
astro
imaging
channel
about
mtf,
and
this
is
the
actual
address
here
and
you
can
type
that
in
and
it
might
be
hard
to
to
snag
that
off
of
the
video
or
even
from
this
slide.
But
if
you
go
to
google
to
go
to
youtube
and
just
search
astro
imaging
channel
mtf,
you
can
pop
up
that
that
presentation
and
that
should
clear
up
a
lot
of
that
stuff.
B
So
just
remember
the
takeaway
here
is
that
once
you
sample
at
more
than
four
point
eight
eight
points
across
the
area
disk,
you
don't
get
any
information
and
here's
what
it
looks
like
this
would
be
a
sensor
array,
that's
sampling
that
airy
disk
and
that's
what
nyquist
sampling
looks
like
at
critical
sampling
interval,
superimposed
on
the
area
disk,
and
this
is
a
2.7
micron
sample
at
f
10.8
again.
I've
just
used
the
celestron
example
and
by
the
way
that
puts
2.05
samples
across
this
central,
the
full
width.
B
Half
max
of
this
central
lobe,
which
is
kind
of
an
interesting
result,
we'll
kind
of
come
back
to
that
later,
but
about
two
samples
across
the
central
lobe
gives
you
the
the
critical
sampling
interval
for
an
arie
disc
and,
as
we'll
see,
this
isn't
really
very
useful
for
ground-based
passive
telescopes.
Okay,
all
you
guys
know
that
the
atmosphere
is
out
there
trying
to
screw
you
up
all
the
time
when
you're
taking
images.
So
this
was
re.
This
is
really
relevant
stuff
for
the
guys
who
designed
hubble.
B
So,
let's
just
talk
about
what
we
can
do
here
when
we,
when
we
sample
holding
the
when
we
hold
the
sampling
interval
constant
in
image
space-
and
here
remember,
the
image
irradiance
varies
directly
as
one
over
the
focal
ratio
squared
the
pixel
area
is
going
to
remain
constant,
in
other
words,
we're
going
to
use
the
same
camera
and
we're
just
going
to
swap
it
between
different
imaging
systems.
So
holding
the
diameter
constant
and
doubling
the
focal
length
decreases,
the
irradiance
and
signal
by
one-fourth.
B
So
if
you
have
your
your
dlsr
and
you
pull
the
the
one
lens
off
and
put
on
a
new
lens,
this
is
what
you're
doing
and
what
that
leads
to
is.
This
is
what
it
looks
like
in
in
the
focal
plane.
We
have
the
same
pixel
spacing,
but
we
end
up
having
different
area
disk
diameters
depending
on
remember
the
focal
ratio.
Remember,
it
only
depends
on
the
focal
ratio,
so
each
pixel
here
when
we
put
on
a
longer
focal
length
lens
on
our
camera,
produces
in
this
case.
B
If
we
double
the
focal
length
we
get.
One
quarter
of
the
signal
so
using
the
same
same
camera
means
that
slower
systems
produce
less
signal
than
faster
systems,
and
even
folks
who
work
in
optics
tend
to
think
this
way.
Okay,
if
I
use
a
faster
system,
I
get
more
more
signal.
Well,
that's
only
true
if
you
use
the
same
sensor,
and
so
you
know
that's,
but
not
so
fast,
I
mean
ultimately,
this
is
why
this
is
called
the
focal
ratio
mish
myth
it's,
because
we're
only
thinking
of
using
the
same
camera
with
different
optics.
B
If,
instead,
we
say,
let's
look
at
holding
the
sampling
constant
in
object
space
rather
than
image
space.
So
now
a
long,
slow
focal
length
system
can
produce
the
same
signal
as
a
fast
system.
If
you
keep
the
angular
sampling
rate
the
same
in
object
space
between
two
different
systems.
So
here
you
can
see
case,
one
we've
got
a
short
focal
length
system
and
here
we've
got
a
case
two
a
long
focal
length
system.
Here
the
signal
is
just
equal
to
the
irradiance
times
the
responsibility
times
x,
squared
which
is
the
size
of
the
pixel
dimension.
B
Here
we've
doubled
the
focal
length.
So
what
happens?
The
irradiance
goes
down,
but
if
we
now
double
the
size
of
the
pixels
to
keep
the
angular
sampling
the
same,
we
get
exactly
the
same
signal
between
these
two
systems
so
for
constant
angular
sampling
focal
ratio
doesn't
matter,
so
it
just
depends
on
whether
we're
holding
sampling
constant
in
image
space
or
whether
we're
holding
sampling
constant
in
object
space.
So,
in
this
case,
focal
ratio
is
all
about
field
size
and
it
has
nothing
to
do
with
signal
strength
if
we
properly
match
our
camera
to
the
system.
B
B
So
this
is
what
that
looks
like
in
the
image
plane
here
we
have
a
short
focal
length,
an
f5
system.
Of
course
that's
going
to
give
a
smaller
psf
in
the
focal
plane,
but
then
we're
going
to
use
smaller
pixels.
Here
we
match
the
pixels
to
the
size
of
the
larger
psf
in
the
f10
system,
with
four
micron
pixels-
and
here
you
know,
we've
matched
this
the
detector
to
the
optical
system
to
maintain
the
same
signal
strength.
So
you
know
using
that
f
10.8
system
in
out
in
new
mexico.
B
If
it's
sampled
properly,
it's
given
me
the
same
signal
as
if
I
had
a
really
fast
system
that
was
properly
sampled
and
here's
an
example
of
that
one
of
the
cool
things
about
the
celestron
telescopes
that
you
guys,
I'm
sure
all
know
about
is
the
you
know.
With
the
14
inch,
we've
got
an
effective
focal
ratio
of
38.50.
B
If
we
use
an
f7
reducer,
we
can
get
that
down
to
2456,
and
if
we
use
a
hyperstar
system
we
can
get
all
the
way
down
to
f
6.75.
So
we've
got
three
different
focal
ratios
and
if
we
use
the
same
camera,
then
indeed
we
get
different
amounts
of
signal.
It's
the
ratio
of
32
to
one
in
one
case
and
13.6
to
1,
in
the
other
case,
with
the
f7.
B
B
So
now,
let's
think
about
seeing,
because
seeing
is
really
the
thing
that
screws
us
all
up
and
it
changes
a
lot
of
stuff
and,
as
you
guys
know,
that
column
of
air
in
front
of
the
telescope
causes
this
time-varying
wavefront
distortion,
and
this
results
in
two
big
effects
for
really
small
scopes.
You
know
in
the
range
of
100
to
130
millimeters.
It
results
in
in
mostly
star
movement,
which
is
tilt
in
the
wavefront
in
larger
telescopes.
Once
you
get
above
this
size,
you
get
tilt
plus
you
get
higher
order.
B
B
If
they're,
coupled
incoherently,
you
can
calculate
the
mtf
of
the
entire
system
by
taking
the
mtf
of
the
sensor,
multiplying
it
by
the
mtf
of
the
optical
system
and
multiplying
that
times
the
mtf
of
the
atmosphere,
okay,
so
turbu.
The
turbulence
theory
developed
by
cola
mcgarv
in
the
1940s
and
freed
in
the
mid
60s
makes
this
a
more
tractable
problem
and
I'm
not
going
to
go
through
all
the
math
here,
but
it's
really
conceptually
very
simple.
B
Basically,
what
we
do
is
we
approximate
that
integrated
star
image,
which
is
a
moffit
by
a
seeing
kernel
using
a
gaussian,
okay
and
the
seeing
induced
gaussian
kernel
has
a
width
of
sigma
in
radians
and
it's
given
by
this
expression
and
where
beta
is
the
stellar,
full
width,
half
max
and
basically
all
you're
doing
here-
and
you
know
this-
this
is
what
cole
mcgorr
off
came
up
with,
but
you
can
do
it
yourself
and
I
did
it
before.
I
read
his
paper.
B
You
basically
just
fit
best,
fit
an
airy
pattern
to
a
gaussian
and
you
can
set
up
the
boundary
conditions
such
that
the
seeing
blur
full
width.
Half
max
is
just
the
airy
full
width,
half
max
with
perfect,
seeing
that's
the
boundary
condition.
So
we
can
use
this
turbulence.
Theory
to
predict
the
full
system
mtf,
and
that
turns
out
to
be
pretty
useful
for
determining.
B
How
do
we
couple?
How
do
we
best
couple
a
camera
to
the
optical
system,
and
so
what
I
did
here
was
kind
of
interesting.
So
first
I
took
that
mtf
curve
and
I
calculated
the
mtf
for
a
diffraction
limited
system
that
has
an
obscuration
ratio
of
39
and
at
this
at
an
f
number
of
f
6.7,
and
that's
this
dashed
line.
That's
what
the
the
20
inch
would
look
like
in
space
and
then
I
did
something
interesting
here.
Instead
of
plotting
this
and
what
you
normally
see
are
normalized
units.
B
I
said:
let's
look
at
this
in
image
space
and
say:
what's
the
required
sampling,
that's
required
in
the
image
plane
in
order
to
sample
each
of
these
spatial
frequencies,
so
to
sample
this
spatial
frequency,
we
need
2.8
samples
per
micron,
okay
or
I'm
sorry
microns
per
sample.
I
said
that
incorrectly
I
apologize.
That's
how
what's
the
spacing
you
need
in
order
to
achieve
perfectly
sample
that
spatial
frequency
now
we
can
also
do
that
in
object
space.
I
don't
really
care
about
object
space.
B
I
just
put
that
in
for
completeness,
but
what
we
can
do,
then,
is
we
can
say,
look
if
we
were
to
calculate
this
just
using
the
obscured
perfect
system
multiply
that
by
the
seeing
values
for
for
a
half
arc.
Second,
one
arc
second
and
two
arc.
Second,
these
are
the
three
curves
we
get.
There's
the
blue
one
is
for
a
half
arc.
Second,
this
is
one
arc
second,
and
that's
two
arc
seconds,
and
you
can
see
what
this
seeing
does
is.
B
It
causes
the
frequency
response
to
fall
rapidly
to
zero
well
before
the
system
does
okay.
So
this
is
what
we're
going
to
see
under
the
atmosphere
so
for
one
arc.
Second,
seeing
in
order
to
properly
sample
that
you
need
a
pixel
spacing
that's
somewhere
in
the
range
of
4.4
to
7.4
microns.
I
just
happen
to
pick
0.1
as
the
cutoff,
so
this
shows
the
range
of
sampling
you
need
to
sample
for
these
three
different,
seeing
conditions
and
that's
just
kind
of
a
starting
point.
B
B
A
B
B
It's
an
imx
455
sensor
that
has
3.76
micron
pixels,
okay,
the
qe
is
higher
than
than
the
fli
the
old
ccd,
but
the
field
of
view
is
a
little
bit
smaller,
there's
a
little
bit
of
a
trade-off
and
I'm
pulling
up
here
a
little
calculation.
I
did
for
the
folks
on
cloudy
nights
and
this
looks
complicated,
but
it's
really
super
simple.
B
This
is
basically
contr
showing
how
to
compare
signal
strength
between
two
systems
of
different
aperture
different
focal
ratio,
different
obscuration,
different
transmission,
different
responsivity,
whatever,
whenever
I
leave
left
oh
and
pixel
size
and
pixel
size.
So
this
is
really
a
very
simple
equation,
even
though
it's
a
little
bit
long
and
it
looks
complicated,
but
what
it
allows
us
to
do
is
to
compare
signal
strengths
between
our
reference
system
and
the
20
inch.
The
the
the
signal
strength
we
call
the
14
inch
100
percent.
B
B
If
we
go
to
cmos
and
we
plug
in
all
the
numbers
for
cmos,
it
looks
like
if
we
use
two
by
two
binning
with
you
know
the
equivalent
of
seven
and
a
half
micron
pixels.
We
get
almost
exactly
the
same
as
we
do
with
the
nine
micron
and
if
we
go
to
the
full
native
one
by
one
binning,
we
only
get
61
percent
of
the
signal,
so
we'd
have
to
use
a
little
bit
more
exposure
to
get
back
to
get
our
signal.
Noise
ratio
back
back.
B
B
The
dashed
line
here
shows
what
we
get
for
the,
and
in
this
case
this
is
the
plane
wave
to
this
is
the
I'm
sorry.
This
is
the
celestron
right
here
showing
what
happens
with
the
nine
micron
pixels
and
what
happens?
Why
does
it
go
to
zero
here?
Well,
that's
the
spatial
frequency
at
which
you
have
a
bright
and
a
dark
fringe
covering
one
single
pixel.
What
happens
to
the
contrast?
B
We
can
do
it
for
the
20
inch
with
the
455,
both
bend
and
unbend,
and
we
can
see
that
as
we
go
to
the
smaller
pixels
okay,
we
begin
to
move
this
response
further
out
and
with.
If
we
use
it
one
by
one
binning,
we
move
the
resp
response
quite
far
out
under
half
arc.
Second
conditions
makes
a
big
difference
and
if
we
were
to
compare
this,
I'm
sorry
so
I
I
said
it
wrong:
this
is
the
20
inch
with
the
9
micron
pixels.
I
I
apologize.
I
originally
said
that
was
the
celestron.
B
B
The
frequency
response
on
the
play
on
the
plane
wave,
the
bigger
telescope
is
worse
than
it
is
on
the
20
inch
under
good,
seeing
conditions.
If
you
actually
sit
here
and
study
this
a
little
bit
under
bad
scene
can
do
well
under
one
and
two
arc.
Second
scene
conditions.
The
difference
is
almost
nothing,
but
when
you
start
looking
at
using
really
small
pixels,
okay,
it's
it.
You
get
a
worse
result
with
the
plane
wave
20
than
you
do
with
the
edge
14..
B
So
the
conclusion
here
is
no
matter
what
and
we
all
know
this
better,
seeing
always
helps
even
a
little
bit.
We
know
that
that's
not
a
surprise,
but
those
charts
show
that,
with
the
9
micron
pixels,
the
20
inch
will
not
produce
a
better
image
quality
than
the
14
inch.
Okay
and
in
excellent
conditions.
It
actually
performs
worse,
so
sending
a
20
inch
to
chile
will
get
sharper
images
with
nine
micron.
Pic
two
get
sharper
images
with
nine
micron
pixels
is
a
waste
of
money.
B
I
mean
I
might
as
well
send
the
the
14
inch
there.
Okay,
so
the
20
inch
does
provide
a
stronger
signal
and
a
wider
feel.
That's
cool!
That's
a
real
benefit.
That's
a
true
benefit
of
the
20
inch,
but
you
don't
need
to
send
it
to
chile
to
get
that
benefit.
I
might
as
well
send
it
to
sro
or
somewhere
else.
B
Small
pixels
only
increase
image
information
under
good.
Seeing
conditions-
and
you
guys
know
this-
I
mean,
but
I
see
people
do
it
all
the
time,
and
I
see
people
talking
about
it
all
the
time
just
use
the
smallest
pixels
you
can
use,
dump
the
the
smallest
pixels
you
can
dump
on
your
your
telescope
and
it'll
all
be
fine.
Well
under
poor
conditions.
Small
pixels
only
decrease
signal
noise.
That's
all
they
do
for
you.
They
don't
give
you
any
more
information.
B
Okay,
so
that's
a
really
big
takeaway!
So
remember
you
know
if
you're
dumping,
one
of
these
new
cmos
cameras
on
a
f10
telescope
under,
say
two
arc-second
conditions,
you're,
making
your
life
harder
as
an
imager.
Okay,
and
then
I
asked
the
question
with
this
system.
What
would
it
take
to
get
a
100
signal
out
of
this
20
inch
compared
to
the
14
inch
that
I'm
used
to
well?
That
would
require
4.8
micron
pixels.
That
would
be
to
me
ideal.
I'd
have
the
same
signal.
B
I'd
get
better
mtf
response,
I
mean
to
me:
that'd
be
a
great
solution,
but
that
sensor
isn't
available
so
but
that's
a
big
takeaway
from
doing
the
mtf
analysis.
Okay,
let's
talk
about
how
we're
going
to
guide
in
the
line
here
and
I
better
hurry
up,
because
we've
got
a
lot
to
talk
about
active
guiding.
You
guys
know
all
about
that
half
arc.
Second,
in
this
case
is
a
diamond
almost
four
and
a
half
miles
4.6
miles.
That's
not
trivial!
B
I'm
not
going
to
talk
to
you
guys
a
lot
about
guiding,
because
because,
if
you're
imagers,
you
know
about
that
you,
you
also
know
about
single
star
guiding
versus
multiple
star
guiding.
Just
before
we
started,
we
talked
a
little
bit
about
auto
covariance,
full
frame
guiding
and
that's
what
I'm
going
to
be
using
on
this
system
and
we're
just
starting
to
test
that
right
now,
up
until
now,
I've
been
doing
single
star
guiding
and
I
get
very
good
performance
out
of
the
system
using
single
star
guiding.
I
also
do
on
access
guiding
using
one
egg.
B
This
is
on-axis
guiding
and
just
very
briefly,
to
re
to
review
that
this
is
the
way
on-axis
guiding
works.
We
have
a
dichroic
beam
splitter,
it
reflects
it
transmits
rather
any
light.
That's
above
seven,
let
me
see
if
I
said
above
700,
nanometers
wavelength
to
a
an
auto
guiding
sensor.
It's
actually
750
nanometers,
like
all
the
visible
light
below
that
wavelength
goes
up
to
our
imaging
sensor
and
back
here.
Our
auto
guide
sensor
is
imaging
in
the
near
infrared
and
we
get.
B
We
can
look
at
the
same
object
that
we're
guiding
off
of
if
we
that
we're
imaging,
if
we
want
to
using
this
arrangement-
and
it
works
extremely
well-
it's
always
easy
to
find
right,
guide
stars
using
this
technique.
I
I
mean
I
very
rarely
struggle
to
find
guide
stars.
That's
extremely
rare!
It's
slightly
less
susceptible
to
seeing
effects.
I
won't
go
into
all
of
that,
but
the
kolmogorov
theory
talks
a
little
bit
about
that.
B
Although
you
only
use
a
very
small
part
of
the
beam
splitter
when
you
use
this
for
each
star,
so
that
turns
out
to
not
be
a
hard
requirement
to
meet,
but
it
is
a
consideration
for
the
guy
who
makes
it,
namely
gaston
over
at
ifi
and
he's
good
at
fit
at
handling
that
the
real
bonus
here
is
that
odin
can
provide
a
focus
signal
and
that's
because
a
converging
beam
excuse
me
through
a
window
produces
astigmatism,
and
this
is
the
exact
same
autofocus
system.
That's
used
in
dvd
players.
It's
been
used
for
years.
B
You
do
have
two
cameras,
the
guide,
camera
and
the
imaging
camera
they
do
have
to
be
very
carefully
co-focused,
but
that
is
a
very,
very
simple
system,
a
very
simple
condition
to
meet
with
very
simple
align
alignment
pre-alignment,
and
it's
only
one
time
you
just
do
it
once
and
the
beauty
of
this
is
I
get
continuous
autofocus
that
runs
in
real
time
with
the
shutter
open.
Now
this
is
most
relevant
when
you
have
a
system
that
is
very,
has
a
very
high
thermal
sensitivity.
B
B
A
lot
of
people
believe
that
they're
rock
solid
you
don't
ever
have
to
re-focus
them,
and
that's
only
true
when
the
temperature
is
pretty
stable,
they
do
change
and
it
depends
on
your
environment.
They
are
less
sensitive,
but
I
do
want
to
emphasize.
They
are
not
completely
insensitive,
and
so
this
business
of
holding
focus
continuously
all
night
long
is
really
a
bonus,
and
by
the
way
this
improves
your
throughput
considerably.
You
don't
have
to
stop
and
take
a
whole
bunch
of
images.
B
Compute
a
v
curve
move
things
around
a
lot
and
then
restart
your
exposure.
This
continuously
focuses
as
you
guide,
and
it's
super
easy
to
use
by
the
way.
This
is
just
a
quick
model
that
I
did
showing
the
geometric
model
of
the
spot
diagrams
through
an
o
nag.
This
is
just
pure
astigmatism,
and
this
is
an
actual
measurement.
This
is
actually
I
bench
tested
the
c14.
Looking
at
what
the
astigmatism
looks
like
through
focus
and
by
the
way
you
get
both
magnitude
and
direction
out
of
this
technique
and
the
sensitivity
is,
is
very
high.
B
The
sensitivity
is
constant
through
focus.
I
said,
linear,
that's
not
quite
right.
It's
constant
through
focus,
unlike
v
curve,
focusing
which
goes
to
lower
sensitivity
as
you
go
through
focus.
This
is
this
is
constant
through
focus
and
you
get
about
plus
or
minus
a
tenth
wave
of
sensitivity
which
is
more
than
sufficient
to
produce
seeing
limited,
focusing,
I
I
like
to
say,
diffraction
limited
focusing,
but
it's
generally
limited
by
seeing
so
I'm
a
huge
proponent
of
one
egg.
This
is
what
it
looks
like
on
the
14
inch.
B
There
is
the
onag
right
there
and,
of
course,
the
imaging
camera,
the
filter
wheel,
the
guide
camera
the
telescope
is
on
an
l
series
mount
which
requires
precision
balance
at
all
times.
So
I
have
this
on
a
rotator,
but
the
rotator
has
to
be
counterbalanced
to
make
sure
that
we
have
good
axial
balance
on
the
mount
when
it
runs.
B
B
A
lot
of
guys
by
the
way,
just
buy
the
scope
and
they
ship
it
to
chile,
and
they
have
the
guys
down
there,
set
it
up
and
that
works
a
lot
of
times,
but
when
it
doesn't
work,
it
is
a
nightmare
and
I'm
aware
of
at
least
one
guy,
and
there
are
probably
more
than
one
guys
who
spent
a
year
trying
to
debug
their
telescope,
the
over
two
7
000
miles
of
distance,
trying
to
get
it
to
work
and
then
ultimately
having
to
ship
it
back
to
the
factory.
It's
a
nightmare.
B
B
This
is
the
exact
same
instrument
that
tested
all
the
optics
from
james
webb
and
the
guys
at
plane.
Wave
set
me
up
so
that
I
could.
I
could
borrow
one
of
these
things.
This
is
you
know
a
very
expensive
thing,
this
this
instrument's
150
000.
So
this
isn't
something
the
average
amateur
can
do,
but
I
could
grab
really
really
good
data.
I
could
show
it's
diffraction
limited
the
diffraction.
This
throw
ratio
is
about
0.85
on
this.
B
It's
not
very
smooth
that
bothered
me
a
little
bit
and
the
biggest
problems
are
right
at
the
edge,
and
I
could
talk
at
great
length
about
why
that
is.
I
know
why
that's
happening,
but
I'm
not
going
to
go
into
that,
but
you
know
my
my
thing
about
this
and
I
want
to
urge
you
guys,
if
you're,
if
you're,
going
to
set
up
in
a
remote
observatory,
be
sure
you
test
it
before
you
send
it
to
a
remote
observatory.
It's
a
nightmare,
fixing
things
in
the
field.
B
The
second
thing
is:
how
am
I
going
to
mount
it?
This
l-series
mount
for
these
plane.
Wave
scopes
comes
and
it
they
stop
right
here
at
the
base
of
this
thing,
and
it's
completely
out
of
balance,
it
requires
putting
it
on
a
pier.
So
I
built
a
roll-out
rig
for
this
thing
out
of
aluminum
I
just
signed
signed
on
to
autocad
and
designed
this
thing,
bought
all
the
parts
and
built
it
and
by
golly
it
works
like
a
champ,
I'm
just
delighted
with
the
way
it
works.
B
It's
rock
solid,
the
whole
scope
by
the
way
weighs
about
700
pounds.
I
almost
rolled
it
over
my
foot
one
day
and
I
can
confirm
it
really
does
weigh
about
that
much
I
mean
it
it.
It
scared
me
when
I
almost
put
it
over
my
foot,
the
electronic
system,
I'm
not
going
to
go
into,
but
there
is
a
big
plan
here
for
running
this
thing
remotely
the
key.
B
The
key
issues
are
the
key
things
that
allow
it
to
be
remote
remotely
operated,
are
number
one,
an
ip
power
switch
that
allows
you
to
turn
on
and
off
various
components
over
the
internet,
and
the
second
thing,
of
course,
is
screen
sharing
on
your
pc.
Those
are
the
two
critical
things
and
I'm
sure
a
lot
of
you
guys
use
that
stuff.
So
it
shouldn't
be
any
mystery,
but
this
thing
is
all
packaged
up.
It's
all
connectorized!
You
can
just
plug
it
in
and
unplug
it
in
seconds.
I
got
it
out
under
the
stars.
B
This
is
a
six
minute,
unguided
exposure
with
the
scope
and
I
zoomed
in
and
gosh.
I
saw
the
expected
diffraction
pattern
and
right
away.
I
saw
diffraction
patterns
at
45
degrees
and
scratched
my
head
and
thought
that
isn't
good,
and
then
I
saw
some
pretty
dark
corners
here
and
I
thought
that
doesn't
look
good,
either
uh-oh.
Well,
the
spikes
were
easy.
I
could
easily
identify
that.
B
It
turns
out
plain
waves,
been
selling
this
for
a
long
time
and
I
think
I'm
as
far
as
I
know,
I'm
the
first
guy
who
ever
noticed
that
there's
a
problem
with
this
shroud.
So
what
I
did
I
solved
it
in
kind
of
a
nice
way.
I
3d
printed
some
a
little
thing
that
just
clips
onto
the
onto
the
structure
and
holds
the
shroud
out
of
the
way
and
that
works
like
gangbusters.
B
B
One
of
the
problems
that
we
have
here
with
those
dark
corners
is
that
calibration,
as
you
guys
know,
corrects
for
three
multiplicative
effects
in
the
signal
cosine
to
the
fourth
which
we
talked
about
vignetting,
which
is
due
to
edges
dust
and
surface
sticks
and
prnu,
which
is
the
photoresponse
non-form
non-uniform
across
the
sensor.
That's
probably
one
of
one
of
the
three
most
important
things
that
you're
correcting
for
in
your
camera.
B
A
lot
of
guys
don't
seem
to
realize
that,
but
that's
a
very
important
thing
and
it's
variations
in
responsivity
between
the
pixels
and
it
looks
like
a
spatial
noise.
That's
proportional
to
the
signal
strength,
so
flats
are
really
critical.
So
when
I
you
know
this
is
the
math
for
flats.
I
think
you
guys
know
that
when
it's
properly
done,
you
have
it
start
with
a
raw
signal,
dark,
dark
edges
and,
and
it
corrects
properly
you
guys
are
images.
You
know
what
that
looks
like
when
I
did
that
with
the
20
inch.
B
Oh
big
troubles,
we
had
big
bright
edges
and
there
are
two
key
reasons
that
calibration
fiat
fails.
Number
one
is
flat
calibration,
does
not
correct
for
additive
signals
such
as
stray
light.
Okay
and
the
flat
calibration
accuracy
decreases
with
increasing
vignetting.
It
cannot
work
with
100
vignetting,
we'll
come
back
to
that
in
just
a
second.
My
first
thought
was
where's
the
stray
light
coming
from,
and
I
looked
in
the
backside
and
sure
enough.
B
I
could
see
right
around
the
baffles
in
this
system
plus
I
could
see
lots
of
veiling
glare
from
from
the
inside
of
the
baffle
system.
The
guys
at
plane,
wave
at
this
point,
were
getting
a
little
tired
of
hearing
from
me,
but
I
called
him
up
and
told
him
about
the
problem
and
they
were.
They
were
great.
They
admitted
immediately
yeah.
We
got
a
problem
with
the
baffles,
so
they
immediately
designed
a
new
set
of
baffles.
For
me
I
and
sent
them
to
me.
B
B
Well,
once
I
started
doing
the
vignetting
and
looking
at
the
corners,
I
was
cutting
out
close
to
50
percent
at
the
corners
and
that
doesn't
account
for
the
optical
vignetting
and
a
few
other
effects,
so
the
corners
are
very
heavily
vignetting
vignetted
and
gaston
was
nice
enough
to
share
his
design
with
me,
and
I
redesigned
one
of
his
parts
and
I
was
able
to
rebuild
a
new
part
that
that
takes
care
of
this
problem.
It
does
require
remounting
the
system
but
it'll
work
on
any
of
the
plane
waves
or
any
fast
system
for
that
matter.
B
Up
to
about
f
6.8,
maybe
a
little
faster.
But
this
is
the
newly
built
parts
that
I
made
showing
the
mounting
parts
for
each
side
of
this
thing.
For
the
onag
and
the
cool
thing
about
this,
is
it
doesn't
change
the
external
dimensions
at
all
and
that's
what
it
looks
like
when
it's
all
put
together
and
aligned-
and
gaston
was
just
wonderful
about
this.
He
helped
me
and
he
helped
realign
this
thing
and
put
it
together
and
bingo
work
like
a
champ.
B
You
can
see
once
I
fix
that
the
vignetting
problem,
as
well
as
the
straight
light
problem
we
went
from
and
by
the
way
this
is
very
heavily
stretched.
You
know
we
get
a
nicely
calibrated
result
and
if
you
ask
me
how
much
vignetting
is
too
much,
that
is
still
a
mystery.
I
I
can't
exact
I
wish
I
could
analytically
solve
this
or
model
it
in
some
way.
I
know
there's
a
way
to
do
it,
but
this
is
a
hard
problem
and
I
don't
really.
I
can't
tell
you
guys
how
much
is
too
much.
B
I
can
tell
you
that
if
you've
got
more
than
50
vignette
in
your
system,
excuse
me
you're,
probably
going
to
have
trouble,
and
I
see
it
all
the
time
on
astroven
I
used
to
see
it
on
cloudy
nights.
This
is
a
problem
that
a
lot
of
people
run
into
so
the
very
first
thing
when
you
get
you
know
bad
vignetting
results.
You
know
a
it's
stray
light
is
the
first
question.
The
second
question
is
how
much
vignetting
you
have
do
you
have
if
it's
more
than
50
you're
likely
you
likely
got
troubles.
B
So
how
do
we
evaluate
the
under
the
sky
performance
and
we're
kind
of
running
out
of
time,
so
I'm
gonna
rush
kind
of
rush
through
this,
but
really
what
I
wanted
to
know
is
how
well
does
this
thing
work
with
all
this
stuff
in
the
beam,
and
you
know
secondary
alignment
all
that
stuff.
How
do
we
do
this?
B
Because
I
you
know-
I
already
measured
this
thing
with
a
150,
000
interferometer
and
probably
a
geez,
a
half
million
dollar
flat
and
a
lot
of
expense,
but
I
still
don't
know
how
it
works
all
put
together.
Okay,
we
could
do
this
with
a
hartman
test
or
shaq
hartman.
We
could
use
a
point
diffraction
interferometer.
We
could
use
ronke
knife
edge,
all
kinds
of
stuff.
These
are
only
qualitative,
all
that
stuff
costs.
B
Money
and
gaston
came
to
me
with
an
idea
quite
a
while
ago,
and
he
and
I
have
been
talking
about
this
for
a
long
time
and-
and
I
have
to
give
him
a
lot
of
credit
for
this,
but
he
and
I
kind
of
worked
on
this
together.
We
ended
up
writing
a
paper
about
it
for
sbie,
but
it's
basically
a
way
to
use
to
to
to
quantify
star
testing
and
the
basic
idea
is
to
generate
some
appropriate.
B
Some
random
use
a
random
number
generator
to
to
generate
some
zerniky
components
that
can
model
a
wavefront,
and
then
we
appetize
that,
according
to
what
we
have
in
the
telescope
and
we
generate
a
monochromatic,
defocus
psf
and
then
we
normalize
it
and
we
compensate
for
coherence
which
involves
what
filter
we're
using
and
that
sort
of
thing.
And
then
we
load
an
end
by
a
neural
network
database
with
n-by-n
pixel
images
and
in
this
case
we're
using
about
600
000
of
these
things.
So
it's
a
huge
amount
of
data.
B
We
put
it
into
a
training
loop
and
we
have
a
figure
of
merit
and
we
train
the
neural
net.
The
the
ai
system
to
recognize
defocus,
star
images
and
when
it
recognizes
the
defocused
star
images
it
just
pulls
up
which
zernikes
were
what
were
involved
in
producing
that
image,
and
we
can
then
compute
a
wave
front,
okay,
which
is
pretty
cool.
So
we
tried
it
and
we
did
this
on
the
bench,
and
these
were
the
first
results
we
did
in
the
field
on
the
20
inch,
and
these
are.
B
This
is
a
raw
image
and
this
is
an
intra-focal
image.
This
is
an
extra
focal
image
using,
as
I
recall,
a
red
filter
and
we
ran
it
through
the
ai
system
and
there
are
the
artificial
images
that
we
got
from
the
ai
system,
and
I
you
know
I
I
present
that
and
say
to
you
those
look
pretty
close.
It
did
that
ai
system
did
an
amazing
job
of
finding
an
appropriately
matched
defocus
star
image
and
the
zernikes
that
went
into
that.
B
If
you
then
go
backwards
and
compute
the
wavefront
from
which
you
can
compute
the
peak
to
valley
the
rms
and
the
strel,
and
then
we
go
back
and
compare
that
to
the
phase
data
now
remember.
This
is
a
single
image
actually
under
the
sky,
with
turbulence
and
look
at
this
this
it,
I
correctly
identified
a
little
bit
of
trefoil.
Unfortunately,
our
color
scales
are
not
the
same,
so
forgive
me,
but
you
can
see
there's
a
lot
of
problem
toward
the
edge
on
this
thing.
B
As
you
can
see
here,
there's
a
lot
of
problem
toward
the
edge
the
most
relevant
numbers
are.
The
rms
here
on
the
interferometer
was
.05065
waves.
Here
we
came
up
with
point
zero:
seven,
five:
zero,
zero,
seven
waves,
it's
only
five
milliwaves
difference.
The
stroll
was
point,
eight,
four,
eight
versus
point:
eight
one,
that's
a
pretty
close
result.
So
this
this
and
again
that
was
a
single
image.
I
think,
if
we
had
averaged
this
and
really
worked
hard,
we
might
get
something
where
I
could
wave
my
hands
and
say
gee.
B
This
is
really
really
good,
but
gee
a
single
image.
It
worked
pretty
well.
So
this
is
a
very
cool
thing.
This
will
be
the
basis
of
what
gaston
will
be
offering
to
the
amateur
community
for
doing
optical
alignment,
because
this
system
can
very
easily
detect
the
aberrations
that
are
involved
when
you
misalign
a
two
mirror
system
so
and
it's
very
sensitive.
So
in
particular
those
of
us
who
do
dso
imaging
this
is
a
great
tool.
I
think
the
planetary
guys
will
just
love
this.
B
I
think
because
they're
always
tweaking
alignment
and
that
sort
of
thing,
but
that's
a
very
interesting
result,
you
know,
and
it
can
be
used
simultaneously
to
image
multiple
stars
over
the
field,
so
we
can
look
at
field
performance
and
all
kinds
of
stuff.
We,
and
so
you
know
I'll,
just
give
you
that
I
think
it's
going
to
be
a
cool
thing
and
finally,
I
did
get
an
image
out
of
the
20
inch
there.
B
It
is
just
from
my
driveway
here
in
bend
before
it
goes
down
to
chile,
we'll
have
it
shipped
in
another,
oh
gosh,
another
month
or
so
here.
So
in
terms
of
imaging,
we'll
just
wrap
it
up,
I
mean
rule
number
one,
and
maybe
you
guys
already
know
this.
I
mean
nothing
is
easy.
At
least
as
easy
as
you
think,
it's
gonna
be,
and
I
don't
care
how
much
optics
you
know,
I
don't
care
how
long
you've
done
this
there's
always
something
and
second,
you
know
always
verify
the
performance.
B
B
When
you
don't
understand
something,
stop
and
figure
out,
what's
going
on
and
pay
attention
to
the
details
that
whole
little
shroud
thing,
you
know
that
that
was
a
little
minor
thing
that
I
just
happened
to
notice,
and
maybe
somebody
else
has
noticed
it
too,
but
I
haven't
heard
it
from
anybody
who
noticed
it,
but
a
lot
of
people
have
said
gee.
I've
noticed
that
thing,
but
I
had
no
idea
what
it
was.
B
Murphy
is
relentless
his
ability
to
to
screw
things
up
goes
on
and
on,
and
you
know
I
do
have
to
say
that
you
have
to
balance
the
urge
to
over
analyze
something
against
trying
it
to
just
see
how
it
works.
You
know
built
relative
to
cost
and
money,
and
you
know
if
you're
building
a
you
know
a
two
billion
dollar
telescope
like
the
gmt.
B
You
need
a
lot
of
analysis,
but
if
you're
building
a
20-inch,
you
know
little
20-inch
telescope,
that's
going
to
go
down
to
chile,
sometimes
just
trying
it
you
know,
is
a
quicker
and
easier
way
to
do
it.
But
having
good
solving
skills
is
a
big
part
of
this
whole
thing
so
I'll
stop
there.
I
apologize.
I
went
just
a
little
bit
over,
but
that's
pretty
much
the
story
of
this
20-inch
telescope
for
you
guys.
So
if
you
have
any
questions,
I'm
happy
to
take
them.
B
A
B
Yeah,
unfortunately,
that's
a
myth,
that's
a
myth.
You
know
the
deconvolution
process
is
not
quite
the
same
as
information
recovery,
but
the
deconvolution
pro
I
mean
unless
you've
got
really
really
you're.
Unless
you're,
you
know
just
seriously
undersampled
in
general,
the
deconvolution
process,
you
know
once
you,
you
apply
the
deconvolution
process,
typically
after
you've
stacked
and
that
will
typically
tend
to
kind
of
fix
a
severe
under
under
sampling
problem,
but
over
sampling
is,
is
not
the
appropriate
response
to
trying
to
get
better
deconvolution
thanks,
yeah.
B
Well,
you,
you
know
the
real
problem,
glenn
is
it's
not
so
there
may
be
some
4.8
and
in
fact
I
right
right
now.
I've
got
I'm
using
a
5.2
or
something
on
a
refractor.
The
problem
is,
I
want
to
use
a
big
sensor.
Okay,
and
so
you
know,
the
problem
is
a
little
bit
more
constrained
than
I
than
I
meant
to
imply
there.
You
know
I
want
a
big
sensor
and
I
want
to
get
something.
That's
you
know
maybe
4.8
ish.
That
would
be
kind
of
ideal.
B
I
mean
if
there
was
a
36
by
36,
millimeter,
cmos
with
4.8
or
or
2.4
that
I
could
bend
to
4.8
again.
You
have
a
bandwidth
problem
for
these
remote
scopes
and
even
with
this,
this
ix,
you
know
455,
sensor,
gosh,
you
know
each
each
image
one
by
one
bend
is
130
megapixel
megabytes,
and
you
know
when
you're
transferring
all
that
stuff
over
the
internet
from
chile
and
you've
taken,
you
know,
say
I
don't
know
40
images
in
a
night
or
whatever
it
happens,
to
be
on
a
long
night.
B
Yeah,
so
you
know
that's
I
mean
that's
a
nice
nice
size
to
bend
to
for
sure.
So
you
know
for
for
f
6.5
and
again
I
you
know,
I
have
to
be
a
little
careful
here.
You
know
with
this
large
aperture
system.
You
know
the
seeing
effects
are
are
always
more
severe.
You
know,
depending
on
what
scope
you
put
that
on
I
mean
that's.
That
should
be
a
really
nice
sensor.
D
E
A
Was
there
somebody
else?
Okay,
I'll
jump
in
you
know,
let's
say
you're
not
sending
a
telescope
down
to
chile,
sure
say
your
imaging
from
the
bay
area
or
a
dark
place
near
the
bay
area.
How
would
your
recommendation
you
know?
I
assume
your
setup
is
way
overkill
for
a
like
that,
where
you
know
we're
seeing
issues
with
some
like
you
know,
you
know,
anyway,
you
get
the
idea.
Well,.
B
Well,
the
advantage
of
this
system-
you
know
in
your
area-
is
that
it
gathers
signal.
Pretty
fast
I
mean
that's
probably
the
biggest
advantage
over
the
14
inch
is
even
with
an
appropriately
sized.
You
know
I
would
always
bend
this
thing.
You
know
if
I
was
putting
this
thing
out
at
dsw.
I
would
never.
B
I
might
put
the
same
sensor
on
it
at
7.4
or
I
might
even
run
it
with
a
16
803,
but
I
would
never
run
it
with
the
small
pixels
and
in
that
case
it
gathers
signal
pretty
darn
fast
compared
to
the
14
inch.
So
you
know
that's
that's
an
advantage
of
the
20
inch,
but
but
the
you
know
the
limitation
there
is
you
don't
get
any
more
image
detail
and
I'm
kind
of
fascinated
by
image
detail.
You
know,
I
see
adam
block's
pictures.
B
You
know
that
he
takes
up
on
mount
lemon
with
that
32
inch
up
there
and
you
know,
he's
got
one
arc.
Second
imaging
up
there
and
you
know
he
gets.
You
know
that
big
telescope
gets
some
pretty
nice
detail
and
I
really
like
the
detail
in
these
in
these
images.
So
that's
a
part
of
my
goal
and
that's
that's.
You
know
maybe
a
little
bit
different
than
than
the
kind
of
goal
you'd
have
to
have
if
you
were
using
it
there
and
by
the
way.
B
D
A
D
A
E
B
Yeah,
so
as
for
location,
I
mean
that's
a
really
good
question
bruce
as
for
location,
if
you,
if
you
look
at
santa
fe
and
draw
a
line
to
to
new
mexi
to
las
vegas
new
mexico,
and
then
you
drop
down
about
from
that
line,
you
drop
down
maybe
40
miles
to
the
south
right
from
the
middle.
That's
about
where
the
observatory
is
it's
out
on
something
called
the
rho
plateau.
It's
about
7
200
feet,
or
so
it's
a
little
over
7
000
feet,
but
you're
right.
B
There
are
a
lot
of
issues
out
there.
When
it's
good.
It's
very
good.
I
mean
it's.
As
I
said,
it's
very,
very
dark
out
there.
The
sky
is
just
spectacular
out
there
you
get
a
little
bit
of
a
light,
dome
from
albuquerque
and
recently
growing
light
dome
from
santa
fe,
but
again
they're
they're,
pretty
dim
and
they're
pretty
low
on
the
horizon.
B
But,
as
you
say,
there
are
issues
you
know
during
the
winter
when
the
humidity
and
they
have
fog
out
there
it
does.
It
does
snow
out
there.
There
are
times
when
the
humidity
is
high
and
they
shut
the
roof.
They're
very
cautious.
To
protect
the
equipment-
and
you
know
that's
a
good
thing,
so
you
know
they
want
to-
they
want
to
make
sure
that
the
telescopes
are
protected
and
humidity
is
one
of
their
parameters.
B
Obviously,
if
rain
is
another
parameter,
they're
very
good
about
keeping
that
roof
closed
because
they
realize
that
there's,
you
know,
there's
a
liability
in
keeping
that
roof
closed
and
they've.
You
know
they.
They
are
dedicated
to
trying
to
protect
your
equipment,
but
during
the
monsoons
you
know,
each
year
is
a
little
different.
This
year
is
disastrous.
I
mean
it
is
absolutely
disastrous.
I've
been
shut
down
for
over
two
months.
I
mean
I
have
to
go
back
and
look
exactly
when
I
stopped
imaging.
B
I
think
I've
had
one
night
of
imaging
this
summer
it,
the
monsoons,
have
been
particularly
strong
this
year,
all
the
way
from
tucson,
where
it's
just
flooding
like
crazy
up
into
new
mexico
and
it
it
can
be
just
a
mess
there.
I
got
caught
in
the
rainstorm
there
at
dsw
one
time
and
the
mud
is
epic
and
the
owner
warned
me.
The
mud
is
really
epic
around
here.
I
didn't
really
believe
him
until
I
got
stuck
in
the
mud
and
I
had
to
get
a
tow
truck
all
the
way
out
there.
B
I
had
a
car
stuck.
I
was
you
know
I
was
sinking
in
the
mud
myself
and
it
was.
It
was
horrifying,
so
it
can
be
challenging
out
there
and
again,
as
I
said
when
I
opened
up
you
know,
the
conditions
can
be
very
good
out
there,
but
the
median
conditions
are
not
sufficient
to
support
a
20
inch.
Now
I
would
say
I
mean
it's
an
excellent
place
for
for
smaller
refractors.
B
The
14-inch
works
really
well
out
there.
I've
got
some
great
images
out
there,
but
you
know
there
are
a
lot
of
nights
where
the
seeing
the
sing
monitor
is
showing
you
know
1.2
arc
seconds,
and
I
can
see
my
guide
star.
You
know
I'm
getting
probably
2.5
arcs
full
width,
half
max
it's.
It
is
nowhere
near
what
the
seeing
monitor
is
showing
when
it's
good.
I
know
it.
I
mean.
B
I
see
the
guide
performance
goes
down
below
0.2
arcs
seconds
rms
and
you
know
the
stars
are
absolutely
I
mean
I
can
measure
on
the
fly
and
I
see
1.4
arc.
Second,
full
width,
half
max.
I
know
when
it's
good
and
it's
not
good
that
often
you
know
I
mean
for
the
14
inch.
You
know
for
smaller
scopes,
it's
no
problem,
but
I
would
not
put
a
20
inch
out
there.
Their
prices
are
reasonable.
The
people
are
just
great.
I
mean
they're
they're
fantastic.
B
D
B
B
A
little
careful.
I
love
santa
fe.
I
love
that
area
of
new
mexico,
but-
and
you
know,
by
the
way
I
mean
sr,
sro
has
has
its
own
issues.
I
mean
every
remote
observatory
has
its
own
issues.
I
you
know
they're,
it's
just
normal
and
I'm
sure
you
know
in
a
year
I'll
be
back
telling
you
about
the
issues
down
in
chile.
Where
you
know
I've,
you
know
it's,
I'm
gonna
find
out.
They
have
issues
too
nothing's,
perfect!
D
A
Back
to
the
technical
part,
I
wanted
to
here's
the
way
I
have
been
thinking
about
it.
So
tell
me:
if
I'm
thinking
about
this
right
or
wrong
so
assume
you
have
a
a
target
that
fits
in.
You
know
your
four-inch
refactor
or
your.
You
know
16-inch
reflector,
whatever
they
the
target
fits
in
both.
So
I
would
think
that
the
number
of
photons
from
that
target
hitting
your
imager,
goes
as
the
area
you
know
ratios
between
the
two
optical
systems
that.
A
B
A
You
know,
given
that
you
optimize
the
you
know
the
pixel
cadence,
you
know
properly,
there's
just
no
ques,
you
know
it,
it
doesn't
depend
on.
F
ratio
is
what
I'm
trying
to
say.
It
depends
on
aperture
area
and
getting
your
camera
sensor
sampled
right.
That's
the
way
I've
been
thinking
about
it
yeah
so.
B
So,
first
time,
let
me
just
say:
you're
not
alone
in
that
and
that
thinking
about
it
that
way,
but
I
would
also
tell
you
that
doesn't
quite
tell
you
the
whole
story.
You
are
gathering
more
photons,
but
what
really
counts
is
the
photon
number
of
photons
per
unit
area,
that's
what
determines
ultimately
the
irradiance
value
and
that
we
measure
that
in
units
of
watts
per
square
meter?
Okay,
so
it's
it's!
B
How
much
power
are
you
dumping
on
the
sensor
per
unit
area
and
so
that
larger,
even
though
those
two
telescopes
may
have,
for
example,
let's
just
start
off
by
saying
they
have
the
same
focal
length?
Okay,
so
so,
as
you
said,
the
image
scale
is
the
same.
The
larger
reflector
okay
will
be
a
faster
system
and
it
will
not
spread
the
photons
over
such
a
big
area,
such
a
big
area
compared
to
the
smaller
telescope,
with
the
much
longer
focal
length
and
much
slower.
B
So
what
you're
really
worried
about
is
not
just
how
many
photons
you're
gathering,
but
it's
the
area
that
they're
spread
over
okay
and
that
determines
their
radiance.
But
the
thing
to
remember
is
that
the
signal,
then,
is
determined
by
how
many
photons
really
you're
intercepting
by
the
sensor
itself.
A
That's
why
I
meant
that
they
both
capture
the
target
so
like.
If
you,
you
would
think
that
you
could
normalize
the
pixels.
You
know
instead
of
thinking
of
irradiances
per
per
meter
squared,
you
could
think
of
it
as
per
pixel,
you
know
watts
per
pixel
and,
and
you
you
should
be
able
to
you-
know,
adjust
the
sensor
such
that
the
larger
you
know,
assuming
you
can
buy
such
a
sensor
such
that
the
larger
aperture
always
dominates
the
smaller
aperture.
B
So
so
I
hear
what
you're
saying
I
I
and
I
understand
how
you're
trying
to
cast
the
problem,
but
let
me
suggest
that
it's
much
easier
to
kind
of
divide
the
problem
into
two
parts.
The
first
part
is
what
is
the
irradiance
in
the
image
plane?
You
know
what
what
really
is
the
irradiance
and
the
irradiance?
We
always
measure
that
in
watts
per
meter
and
a
faster
system
always
produces
higher
watts
per
square
meter
than
a
slower
system.
B
So
we
know
that
right,
but
the
signal
itself
it
is
deter
determines
is
the
important
thing
and
that's
determined
by
how
we
divide
up
that
watts
per
meter,
so
you're
right,
but
the
focal
ratio
myth
is
that
a
fast
system
is
always
better
than
a
slow
system.
Well,
that's
true!
In
irradiance,
okay,
that's
true
when
you
talk
about
the
raw
irradiance
in
the
image
plane,
but
that's
not
true.
When
you
talk
about
the
signal
that
actually
comes
out
of
the
sensor,
maybe
that's
the
way
to
say.
A
B
Yeah
yeah,
that's
right!
That's
right!
I
just
I.
I
tend
to
shy
away
from
the
notion
that,
and
it's
a
true
notion,
it's
absolutely
true-
that
the
bigger
telescope's
gathering
more
photons-
that's
absolutely
true,
but
but
the
I
tend
to
shy
away
from
starting
at
that
point,
because
I
find
that
it
gets.
It
confuses
the
issue
because
really
what
really
counts
is
how
are
all
those
photons
spread
out
in
the
image
plane?
That's
what
ultimately
counts
in
terms
of
irradiance.
A
B
Fine,
well,
it's
been
a
pleasure
talking
with
you
guys
and-
and
I
appreciate
the
opportunity
to
talk
to
you-
it's
been
great.
Hopefully
we'll
see
you
again,
maybe
at
some
meeting
maybe
aic
in
this
fall.
This
fall
hopefully
they'll
hold
the
meeting
but
I'll
be
at
aic.
If
they
hold
the
meeting,
didn't
they
move
it
to
2022
already.