►
From YouTube: SimPEG meeting June 3, 2020
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
B
B
I,
don't
know
what
to
say
things:
I,
I
think
I.
Finally,
some
sort
of
survey
implemented
that
I'm
happy
with
and
in
the
end
I
went
under
the
hood
with
xra
I've
tried
a
lot
of
different
things
and
wrote
code
and
deleted
code
over
and
over
again,
because
somehow
never
you
happy
and
in
the
end,
I
sort
of
wanted
to
avoid
the
X
right
so
now
for
some
reasons,
but
now
yeah
I
just
jumped
in
with
it.
Maybe
it
makes
things
easier
down
the
line
at
the
beginning,
I
didn't
find
it
very
easy.
B
B
Not
much
not
much
I
started
with
the
thing,
and
if
you
have
time
at
the
end,
maybe
we
could
spend
a
few
minutes
about
to
transform
how
you
guys
think
you
might
have
time
when
or
even
if
he
maybe
could
meet
Friday
and
look
with
like
pain
or
Miguel
to
have
something
like
beforehand
a
meeting
what
he
expects
for
it,
because
by
the
time
you
will
be
online
on
Saturday,
most
of
the
day
is
gone
for
us.
So
we're
here
so
it
might
be
good
to
bang
our
heads
together
beforehand.
If
there
is
interest
sure.
B
C
B
B
Although
the
second
weekend,
because
of
the
different
time
zones
Sunday,
is
basically
just
presenting,
mostly
because
by
the
time
you
get
online
on
Sunday,
it's
basically
the
weekend
is
over
for
the
people
here,
so
it
will
be.
It
will
have
wrapped
up.
So
it's
really
the
first
weekend,
and
then
people
who
want
to
spend
some
time
during
the
week
want
or
can
spend
some
time
during
the
week
and
then
the
Saturday
and
then
it's
a
wrap
up.
I
would
say
so.
B
I
was
thinking
this
Friday
beforehand,
so
that
they
would
have
a
bit
a
clearer
idea.
What
we're
gonna
do
one.
My
personal
idea
is
that
it's
quite
open.
So
if
there
is
someone
with
quite
a
couple
signed
up
in
the
slack
channel
and
if
there
is
someone
who
just
wants
to
make
a
notebook
and
integrate
X
with
Y
I
think
that's
perfectly
fine
and
as
Miguel
is
around
is
around,
and
some
of
you
guys
are
around.
B
They
should
get
the
help
to
connect
the
different
dots,
but
it
looks
like
also
Miguel,
and
they
would
be
interested
in
doing
something
more
long-term
ish
with
connecting
these
different
things.
So,
for
this
week
probably
should
maybe
have
a
talk
and
then
I
think
it'll
be
good.
If
you
could
have
that
Friday,
so
we
would
have
two
full
weekend
for
the
people
who
I
have
time
and
our
interest
so
I
don't
know
I
like
9
10
o'clock,
you
time
what?
What
time
do
you
have
now.
B
D
C
C
B
A
B
No
I
I
also
I
expected.
We
probably
people
have
no
time
at
all
weekend
if
they
still
have
two
kids
homeland,
but
maybe
they
can
chip
in
an
hour
each
day.
During
the
week
err
I
mean
it's
quite
proud.
Two
topics
over
yeah.
If
you
can't
have
an
eye
on
slack
and
on
that
channel,
and
if
there
are
folks
who
try
to
do
something
its
impact,
that
would
be
helpful,
I
guess
already
yeah.
F
B
A
Thought
I
had
on
that.
So
for
folks
who
are
in
this
swung
slack
if
you're
not,
you
should
be
first
of
all
and
then
what
I
was
thinking
is.
You
know,
there's
a
lot
of
folks
who
put
like
the
name
of
where
they
work
after
their
name,
I
thought
actually
for
this,
that
for
the
tutorials
in
the
hacking
it
might
be
constructive
if
we
all
just
put
like
simpe
behind
behind
our
names,
so
that
if
folks
have
questions
or
things
like
that,
it's
super
obvious
that,
like
yes,
these
people
know
no
same
thing
again.
A
A
A
So
that's
actually
these
well
run
this
topic
to
you.
That's
one
thing:
I
would
appreciate
for
the
tutorial
that
I'm
doing,
which
is
on
Tuesday
from
11
specific
till
QP
ins.
If
some
folks
could
just
be
on
slack,
there
will
be
a
specific
T,
20,
syn,
peg
channel
and
so
I'll
trying
to
add
everyone.
A
But
if
you
can
also
add
yourselves
and
help
answer
questions
there
and
then
also
potentially
just
keep
an
eye
out
for
if
there
are
questions
that
are
actually
worth
sort
of
surfacing
and
me
speaking
to
directly,
because
I
will
try
and
keep
an
eye
on
slack.
But
it's
also
it
could
be
a
bit
tricky
depending
on
how
many
questions
come
in
and
things
like
that,
and
so,
if
folks
here
could
just
flag
and
like
ping
me
personally
and
just
say:
hey
Lindsay.
This
was
a
really
good
question.
A
G
H
G
Mean
I'm
just
working
away
at
the
1d
problem,
so
I'm
more
or
less
figured
out
the
kind
of
local
1d
Ford
modelling
and
inversion
and
kind
of
set
up
a
more
simple
like
source
and
receiver
class,
and
really
made
it
quite
generalizable
and
played
around
with
waveforms
and
things
like
that.
So
I'm
kind
of
wrapping
up
the
kind
of
a
single
sounding
and
I'll
be
moving
to
the
stitched,
the
stitched
1d
problem
and
trying
to
get
that
to
work.
So
I
think
once
I
get
to
that
point.
G
A
C
A
I
accepted
it
Joe
do
we
want
to
I
guess
we
don't
really
need
to
bump
our
version
on
this
there'll
be
other
stuff
coming
in.
C
A
Okay:
ninety
submerge
yeah.
A
C
Yeah
other
than
that
I'm
still
working
on
you
know:
I
had
like
gooeys
to
invert,
one
D
code
and
I
grabbed
mag
and
now
I'm
kind
of
trying
to
consolidate
everything
in
a
single
single
GUI
and
now
that
can
make
me
think
of,
like
all
the
other
methods
that
are
gonna
come,
and
you
know
how
to
organize
master
gooeys
and
drivers
and
all
that
stuff.
So
that's
what
them
I've
been
working
on
recently
and
yeah
I
guess
transform.
What
do
you
need
from
me
like?
So
we
inverted
Raglan
right.
It
looks
alright.
C
A
A
But
if
you
do
have
a
notebook
that
like
we
could
even
distribute
as
it
part
of
that,
that
would
be
great,
but
if
not
that's
also
fine
I
mean
I.
Think
like
so
one
of
the
things
that
I
was
thinking
would
be
pretty
cool.
To
do
is
we're
gonna,
distribute
all
the
lines
of
the
century
data,
but
only
walk
through.
A
One
of
them
is
basically
to
then
ask
if
folks
want
to
run
inversions
on
all
of
these
and
send
us
like
a
link
to
their
notebook
and
their
inversion
results,
and
we
can
sort
of
get
a
whole
suite
of
different
models
from
everyone
who's
participated,
and
so
we
could
also
do
that
with
Raglan.
If
two
toriel
to
go
on
but
I,
don't
think
it's.
It's
not
super
critical.
A
C
A
A
D
Think
good
things
are
like
yeah,
so
thanks
for
your
review
on
such
don't
paper,
so
yeah
I
went
through
that
like
have
a
working
draft
of
the
of
the
tasty
signs
that
I
need
to
reread,
and
then
we
and
them
I,
just
I.
Recently,
you've
probably
seen
every
submitted
the
dkc
paper
too,
so
not
easy
working
on
two
papers
and
wanted
this
one.
At
the
same
time,.
A
I
Yeah
actually
I've
been
up
to
quite
a
few
things:
lots
of
memory
management,
so
I've
had
to
slow
down
a
little
bit
of
code
to
just
try
and
get
things
done.
I
found
some
interesting
things
with
party.
So
though,
I
don't
know
if
I
can
share
my
screen,
I
have
the
problem
up
and
I
can
just
I
can
share
screen.
Okay,.
I
I
C
I
It
was
kind
of
unfortunate
as
it
kept
happening.
He
took
me
a
while
to
debug
to
find
it
was
just
that
so
then
I
thought,
oh
well,
I'll
just
throw
the
chunking
at
it
that
we
introduced
their
dog
and
now
it
works.
So
if
I
there
seems
to
be
a
limit,
I,
don't
know
or
something
with
party
so
where
it
tries
to
check
that
matrix
and
if
it's
too
big
it
just
throws
an
error.
I
think.
B
I
I
I
H
J
J
C
I
H
Oh
sorry
really
is
my
microphone
like
this
is
a
good
use
case
for
in
place
solutions,
because,
as
long
as
you
can
like
sort
of
store
it
on
the
right-hand
side,
you
can
overwrite
that
as
you're
doing
the
solution,
so
you
can
probably
have
any
handle
larger
and
larger
right-hand
sides,
because,
right
now
what
climate
solver
is
doing
is
it
creates
it
it
like,
by
default.
It'll
just
create
another
array,
exactly
the
same
size
as
your
right-hand
side.
H
H
There's
definitely
a
lot
of
issues
with
time
at
solver
and
or
the
PI
entail
implementation
that
it
uses,
because
it
doesn't
throw
many
good
errors.
It
doesn't
give
you
many
good
air
warnings,
which
is
something
that
I've
talked
about
before
trying
to
get
a
better
nko,
rep
or
party,
so
rapper
entails.
We
can
handle
those
type
of
errors
or
catch
them
and
then
do
a
few
more
operations
that
we
that
we
can
do
and
take
advantage
of
some
things
like
that
in
place.
Solutions
and
whatnot.
I
I
I'm
not
even
found,
if
I'm
trying
to
save
those
each
of
those
matrices
to
desk
I
still
run
out
of
or
if
I'm
just
trying.
To
sum
all
those
into
a
desk
array
I'm
still
hitting
over
the
limit
to
on
RAM,
so
I
think
if
you're
yeah
doing
that
in
parallel,
it's
just
bringing
them
voting
them
all
in
and
then
overloading
my
RAM.
So
right
now,
I'm,
just
yeah
I'm,
just
saving
everything
to
disk
and
then
I'm
just
gonna
pull
everything
iteratively.
I
I
Yeah.
That's
ten
I've
got
about
a
frequency
for
all
components:
all
stations
at
about
eight
hours,
that's
about
as
fast
as
I
can
get
it.
It's
I
know
I'm
parallelizing
over
station,
maybe
at
that
size
doing
it
over
component
or
something
would
be
better
at
state.
You
started
testing
it
at
small-scale
if
I
found
that
going
over
receivers
was
the
best
but
yeah.
I
That's
a
big
problem,
and
almost
after
my
or
might
have
to
start
dividing
it
in
by
frequency
and
sending
that
to
a
node
like
a
node
server
and
so
for
frequency
and
parallelize
over
that,
and
then
stop
underneath
those
yeah
something
I
might
get
into
in
the
next.
A
little
bit
here,
because
yeah
take
in.
C
I
C
Do
you
started
noticing,
like
you,
I'm
getting
I'm
working
on
this,
because
a
bunch
of
people
are
starting
to,
you
know,
run
a
work.
They
want
to
run
like
huge
terabyte
problems
like
you're
best
for
some
reason,
start
idling
after
certain
level
like
you're
all
excused
are
firing
up
in
them.
You
have
a
idle
time.
You
know,
I
can
go
for
like
a
few
few
seconds.
I've.
C
I
I've
just
been
noticing
that
here
in
the
last
week,
yeah
all
of
a
sudden
it
just
stops
and
I.
Don't
know.
I
was
thinking
and
maybe
it
when
it's
it's
happening
in
the
operation,
to
dare
to
desire.
I
thought,
maybe
there's
something
in
there.
That's
just
single-threaded,
but
I,
don't
know
I
just
series
I
write
twos
are
you
can
see
the
blip
when
it
writes
and
it's
so
quick,
so
I
can't
see
why
it's
sitting
there
just
spinning.
C
C
Yeah
I
do
try
to
use
it,
but
the
thing
is
that,
like
yeah,
we
don't
get
the
same
paralyzation
if
we
run
it
in
a
notebook
versus
run
it
and
in
a
script
like
yeah,
we
get
like
overhead
just
using
the
graphs
or
it's
kind
of
hard
to
it's.
Not
it's
not
super
easy,
but
yeah.
That's
one
option
for
sure.
I
Yeah,
look
at
that
yeah
I've
just
been
working
on
that
GUI
I!
Pretty
much.
Have
it
done,
it's
pretty
slick,
so
I
have
a
whole
bunch
of
like
I,
guess:
DC,
an
IP
server,
server
code
kind
of
thing
so
that
you
send
it
just
a
JSON
file.
Is
the
config
file
and
yeah
it'll
run
locally
on
another
another
resource.
I
Yeah.
Actually
I
was
wondering
if
I
can
get
some
simpang
icons
so
that
I
can
put
into
the
app
because
we're
eventually
thinking
that
we
could
like
the
clients
will
have-
and
we
could
push-
will
be
like
Oh
new
new
inversion,
ready
and
it'll
be
like
a
little
bag
icon
just
to
keep
everything
there.
So
everybody
knows
what's
what
it's
powered
by
and
whatnot
awesome
yeah.
A
I
I
I
All
right
yeah,
so
it's
simple
login
your
sessions
are
all
stored
on
on
a
cloud
for
save
whatever
you
want
to
call
it
yeah.
It's
simple
things.
So
of
anything
all
those
are
there.
You
want
to
start
a
new
session.
You
can
just
name
your
sessions
here
in
version
type,
but
so
I
thought,
like
maybe
a
little
less
in
peg
icons
here
it
would
be
cool
just
powered
by
will
have
forward
and
then
processing
we're
just
gonna
have
a
lot
of
our
stuff
live
under
here.
I
So
it's
easy
to
pass
around
under
our
hood
here,
conversion
types
and
then
just
a
main
interface
for
just
unconstrained,
inversion,
kind
of
thing:
we're
not
looking
to
be
too
fancy
so
something
quick
that
are
even
maybe
in
the
field.
Some
guys
can
check
their
data.
If
we
have
bad
connection
and
they
to
do
it
from
there,
they
can
process
the
data
and
send
it
into
here.
You
can
load
it
locally
or
you
can
have
there's
going
to
be
a
path
on
our
file
server,
so
you
don't
have
to
load
yeah,
set
all
your
options.
I
I
tried
to
generalize
it
enough,
but
I
guess
over
in
the
additional
inputs
is
kind
of
where,
but
we're
have
your
conductivity
model
gives
not.
Everything
will
have
that
so
I
just
kind
of
threw
something
through
this
until
I
got
mismatch
area
yeah,
then
you
can
choose
our
resource
right
now.
We
just
have
four
resources
available.
Our
big
machines
are,
of
course,
are
named
real
names
yeah.
So
when
you
can
submit
your
in
version
and
it
will
update
in
real
time,
but
usually
ours
are
so
long.
I
To
get
rid
of
there,
but
you
can
go
back
to
your
previous
session
say
you
ran
in
it
or
you
started
the
inversion
yesterday
and
you
want
to
come
check
what
how
it's
doing
you
can
request
the
inversion
log
and
then
it'll
yeah,
pretty
much
outputs
what
you
guys
have
in
that
directive
output.
It's
pretty
much!
This
looks
for
that
file
and
then
sends
it
back
to
you.
Yeah
tells
the
inversion
is
complete,
so
then
we
can
get
the
model
or
we
can
choose
one
of
the
iterations
get
our
residuals
like
a
deist.
C
I
A
secret
sauce:
well,
it's
pretty
much
it's
just
more
like
what
the
data
wise
error
on
the
data,
so
like
stability
of
the
on
time,
how
well
the
vs
get
the
cocoa
curves,
actually
soggy,
stretch
exponential!
That's
what
we
measure
our
error
with
see
how
well
those
good
yeah
so
I
haven't
that
all
that
is
written
in
Python
2.
So
if
we
ever
need
like
interface,
stuff
I've
got
a
lot
of
code
written
because
everything's
prototypes
in
Python,
so
yeah.
I
C
J
I
C
Working
on
on
the
my
gooeys,
a
if,
when
you
get
to
the
GA
side,
I
started
using
the
comments,
you
know
create
a
folder
and
put
comments,
so
you
can
stack
your
JSON
file
right
there,
instead
of
it
being
on
file.
Oh
pretty
useful,
so
I
can
look
at
it.
Yeah
that'd
be
cool
and
say
about
your
iterations
s,
comments
right
on
you
and
version
folder.
So
then
you
don't
need
to.
We
got
the
print.
You
can
just
store
it
all
as
string
and
then
reload
that
does
it
JSON
out
there
yeah.
A
Meant
done
actually
once
you've
got
it
on
pi
PI,
it's
actually
pretty
straightforward
to
get
things
on
kondeh
Forge!
You
can
actually
build
the
kind
of
Forge
package
from
pipeline.
So
if
you
have
questions
and
with
that
feel
free
to
obtain
cuz,
it's
nice
nice
to
have
those
options.
C
J
Nothing
particular
I
worked
a
little
bit
on
the
century
century
deposit
data,
so
it
was
kind
of
fun
to
revisit.
Like
you
know,
ten
years
ago,
historical
data,
but
yeah
I,
was
surprised.
I
regard
we
can
get
pretty
similar
result
from
20
years
ago,
so
I
had
to
kind
of
fine-tune
the
problem
at
the
parameters.
J
War,
but
without
pretty
similar
sort
of
like
old
dcpip
judy
was
old
town,
so
I
think
Lindsay
will
use
that
in
the
transform
workshop
and
yeah
I
also
tested
with
thumbs
sports
inversion,
not
sure
it's
like
not
sure
that
I
can
have
an
image,
but
that
I
can
show
you
guys
later,
but
we
can
actually
get
pretty
good
match
with
the
defending
they
were
trying
to
image
so
yeah
I'm,
actually
pretty
it
was
actually
pretty
fun
and
we
ended
up
inverting
in
3d.
So,
potentially
you
do.
J
J
Like
every
stupid,
it
was
not
particularly
useful.
I
also
did
that
I
also
did
this
first
inversion.
It
kind
of
fight
so
like
you
have
to
carefully
choose
the
background.
Resistivity,
but
I
think
that,
like
kind
of
just
doing
a
first
step,
smooth
inversion
and
picking
maximum
likelihood
value
of
your
distribution,
it's
actually.
D
J
Other
when
we
were
working
with
this
example
note,
but
we
found
like
a
number
of
KY
was
a
bit
of
a
problem.
So
I
think
there
was
a
bit
of
accuracy,
I'm,
not
sure
what
happened,
because
I
sort
of
test
it
before
like
about
sort
of
5%
ish
error,
but
it
was
actually
just
showing
us
more
than
5%
like
10%,
especially
at
the
near
surface
like
it
like
that
small
n
spacings
I
had
to
increase
the
number
of
KY
in
the
two
and
a
half
so
I'm,
not
sure
Peter.
J
You
probably
have
a
better
idea
so
like
this
is
same
like
Henkel
transform
types,
a
problem
like
it
basic
functions
to
cosine.
So
what
we
did
is
very
good
for
us
like
a
trapezoidal,
so
we
just
discretize
in
like
I,
don't
really
and
then
just
did
that
trapezoidal.
So
if
you,
you,
probably
have
much
better
idea
that
how
we
can
kind
of
optimize
that
process
so
it'll
be
kind
of
interesting
to
hear
your.
B
B
J
A
it's
a
spatial
like
Henkel
transform,
but
it's
a
space
domain
wha
like
KY
to
Y
and
it's
a
cosine
transform
so
I
think
the
one
easy
way
as
you
do
like
you
like.
Basically
before
we
did
the
15k
Y
value
evaluation,
but
we
can
actually
do
a
sprint
spline
interpolation
in
between
and
I
think
as
you
as
you
do
so
I
think.
That's
we
better
cuz
like
now
what
we
had
to
do.
We
had
to
just
increase
the
number
of
ky.
They
were
computing
which
will
increase
our
computation
cost
so.
B
J
H
B
H
H
H
H
H
So
we're
doing
it
right
now
from
like
the
wave
numbers
are
between
like
10
to
the
minus
4,
to
10
to
the
1.
So
if
you
go
a
little
bit
lower,
it
gets.
If
you
go
10
and
minus
5
and
10
to
the
1,
it
does
a
little
bit
better
too
long
offsets.
It
requires
a
little
bit
more
somewhat
like
the
weights
I
mean.
Ideally,
the
weights
would
be
dependent
on
like
the
source,
receiver
offsets,
but
it's
really
hard
to
do
that.
It's
much
easier
just
to
like
throw
it
all
the
ones.
H
I
give
everything
the
same
way.
I
think
my
point
is
at
least
using
the
Gauss
quadrature.
It's
a
lot
more
accurate
than
the
quite
well
and
then
I
think
I
put
in
there
I'd
like
to,
but
we'll
just.
We
can
just
expose
that
NK
wise
for
people
to
play
with
it
see
if
they're
pork
that
needs
to
get
changed,
a
candle
will
leave
it
defaulted.
I,
don't
know
2015
15.
It
depends
on
a
problem.
I.
F
Do
you
know
Joe
if
or
sucky
was
there
ever
an
implementation
of?
You
know
the
old
filter
ways
of
working
with
this?
You
know
with
the
gauche
filters
or
the
anderson
filters.
I
think
the
initial
dcpip
may
have
been
coated
up
that
way
and
now
we
can
get
by
with
I,
don't
know
11
or
something
like
that
could
work
for
a
variety
of
circumstances,
I'm,
not
sure
whether
that
is
just
like
so
ancient
technology
that
nobody
looks
at
it
anymore.
If
that
would.
J
I
think
that
was
actually
what
I
was
exactly
asking
to
Peter.
So
Peter
has
a
lot
of
were
like
a
sort
of
flexibility
on
that
part.
That
kind
of
even
like
designing
of
filters
and
stuff
so
I
think
he's
got
a
lot
of
tools
Stud.
We
can
do
that,
but
what
I
wasn't
sure
like
you
may
not
have
exactly
kind
of
like
felt
exact
awful
that
we
would
like
to
use.
So
we
probably
do
something.
B
J
F
F
F
B
It's
the
same.
It's
the
same.
We
use
for
our
furry
and
hunk
of
transforms
just
in
the
beginning
of
the
filters
exactly
they
did.
They
did
when
there
to
slumber
Shea,
filtering
and
backwards
and
and
all
these
type
of
things,
but
6
and
10
and
11
and
15
points
yeah
I.
Don't
think
that
it's
done
much
these
days,
it's
yeah,
but
it
would
be
interesting.
I,
but
I
preserve
I
personally
only
have
the
experience
with
the
hand,
column
and
Fourier
transforms
not
with
any
type
conversion,
but
it's
the
same
principle.
In
the
end.
H
H
A
H
A
Yes,
oh
I
think
it's
coming
together.
I
can
give
my
beep
well.
Joe
briefly
showed
the
notebook
and
I
can
give
a
quick
overview
of
like
what
I'm
thinking
and
then
I'll
definitely
appreciate
feedback
and
probably
be
sending
this
out
to
all
of
you
for
a
review.
One
of
the
things
I
would
like
oh
wait.
I'm
on
the
wrong
person.
Let
me
fix
that
before.
A
A
H
A
I
know:
that's
all
right.
You
are
always
welcome
to
okay,
so
we'll
give
a
bit
of
background
on
the
century,
deposit
and
so
Doug
is
on
and
then
wanted
to
give
just
a
bit
of
motivation.
So
I
grabbed
the
IP
inversion
results
from
the
Lenten
paper.
That
was
done
with
the
UPC
code
and
basically
say
okay
like
can
we
actually
replicate
those
results
and
maybe
update
the
color
bar,
so
we'll
walk
through
the
dependencies.
A
I'll
talk
just
briefly
about
sort
of
some
of
the
things
that
are
in
this
impact
ecosystem
and
how
stuff
is
broken
down
like
disco
ties
and
I'm
at
solver
insolence,
so
we'll
go
ahead
and
import
the
things
and
set
a
nice
font
size
I'll
talk
a
little
bit
about
what
the
data
are
so
we're
doing.
A
So
then
I
wanted
to
show
like
actually
just
what
is
the
data
file.
So
now
we're
actually
just
looking
at
the
data
file
and
we
can
see
there's
a
comment
line
then
there's
the
number
of
sources.
This
are
flags
and
then
we've
got
the
a
electrode,
B
electrode,
a
number
of
receivers
and
then
M
and
datum
and
standard
deviation,
and
that's
just
repeated
so
then
I
wrote
a
little
function
to
actually
go
in
and
load
those
data
and
I'll
walk
through
this
I
mean
so
I
know.
A
We've
got
lots
of
this,
basically
in
the
IOD
C,
but
I
thought
it
was
constructed
to
actually
just
like
start
from
scratch
and
like
here's,
a
data
file
and
we're
gonna
load,
and
so
what
this
function
is
gonna
do
is
then,
give
us
back
a
dictionary,
so
we'll
get
the
a
locations,
B
locations
and
locations
and
the
observed
data,
standard
deviations
and
the
number
of
sources.
And
then,
as
we
run
it
I
just
printed
out
some
stuff
to
sort
of
sanity,
check
that
what
we're
doing
makes
sense.
A
So
but
then,
if
we
go
and
actually
look
at
sort
of
what's
in
that
dictionary,
I've
got
an
umpire
array
of
the
a
locations
and
B
locations,
and
then
the
M
and
an
absurd
data
and
standard
deviations
are
all
lists,
and
so
the
way
that
that's
all
organized
is
here.
We've
got
like
you
know,
a
single,
a
location
for
each
of
the
sources.
But
then
we
look,
for
example,
if
we
look
at
the
end
locations
now,
each
entry
in
the
list
is
an
array
of
locate,
end
locations
associated
with
that
given
source.
A
So
everything
is
in
the
order
of
the
sources
yeah
we
can
see.
We've
got
27,
ok,
so
there
we
go
and
I'll.
So
what
the
sim
page
survey
is
like,
so
talking
about
creating
a
receiver,
so
we've
got
a
dipole,
that's
got
M
locations
and
end
locations,
so
you'll
see
actually
in
simply
Joe
and
I
and
Sagi
up
and
having
some
discussions
on
a
few
of
the
issues.
Just
about
naming
and
things
like
that.
A
So
this
is
now
a
slightly
more
pet
name
than
the
capital
M,
which
we
had
before
it'll
still
all
be
backwards
compatible,
but
nicer
for
the
tutorial
to
show
nice
names,
I
build
up
the
receiver
list,
build
up
the
source
and
then
build
up.
The
survey
so,
of
course
walk
through
what
that
looks
like
and
then
I
talked
a
bit
about
the
data
class
and
basically
just
say
you
know
we're
we're
basically
wrapping
or
unwrapping
everything
the
same
way
as
what
we
do
with
the
survey.
A
Then
we'll
plot
up
the
pseudo
section.
So
that's
voltages
I've
got
a
little
widget
to
just
show
how
pseudo
sections
are
built
and
so
the
way
that
we're
actually
visualizing
the
data
and
then
show
the
apparent
resistivity.
A
Don't
go
through.
We
can
look
at
a
histogram
of
the
data
and
from
that
use
that,
as
a
way
to
estimate
a
model
there's
a
few
different
ways,
we
can
think
of
doing
that.
We
could
do
the
mean
if
we
look
in
logarithmic
clear
in
this
case,
I
chose
to
do
the
median
or
you
could,
alternatively
go
and
pick
the
maximum
likelihood.
We
can
take
a
look.
G
A
G
A
Okay,
so
that
takes
us
through
just
like
what
are
the
data
and
then
we'll
set
up
the
forward
simulation.
So
I'll
talk
through
in
this
case,
and
this
is
where
it
this
is
in
one
particular
area
where
I'd
really
appreciate
thoughts
incomplete
because
I
don't
want
to
go
too
far
into
like
finite
volume
and
discretizing
stuff,
but
want
to
give
just
enough
detail.
So
folks
know
what
are
some
of
the
important
decision
factors
in
designing
it
mesh.
So,
in
particular,
boundary
conditions
is
an
important
one,
but
also
when
visualizing
stuff.
A
If
we
visualize
fields
or
flexes,
you
need
to
know
where
they
live,
so
we'll
go
through
we'll
design.
A
mesh
and
I'll
talk
through
you
know
and
I
will
talk
through.
Basically,
just
some
of
those
decision
points
work.
A
A
So
we'll
talk
through
just
like
what
what
some
of
those
things
are
set
up
and
run
a
simple
forward
simulation
for
the
background
model,
and
then
this
is
where
that
discussion
came
about
with
understanding
the
pseudo
section.
What
was
going
on
there
and
show
basically
that
that's
a
ways
of
checking
your
mesh
design
so
right
now
we're
within
basically
five
percent
error
using
the
thirty
case,
and
so
if
we've
got
five
four
five
percent
or
greater
errors
on
our
data,
then
this
is
good
enough.
A
A
Just
so
folks
can
get
a
sense
of
currents
and
things
like
that,
because
I
think
that
that
was
actually
one
of
the
an
extremely
constructive
thing
and
folks
got
pretty
excited
when
we
gave
the
VCG
s,
talk
that
you
can
actually
look
at
the
currents
and
the
earth,
so
I
was
thinking
to
show
some
of
that
and
then
go
through
and
start
from
a
synthetic
model
and
walk
through
some
of
the
important
choices
in
inversion
I'm.
A
So
talking
about
the
alphas,
the
beta
are
misfits
and
uncertainties
assignments,
and
things
like
that
to
give
scent
to
give
folks
a
chance
to
just
like
play
around
with
the
synthetic
conversion,
for
that
I
was
thinking.
Basically,
just
maybe
like
a
layer
and
two
blocks
might
be,
might
be
useful.
You
know
something
complex
enough
that
there's
a
bit
of
interest
in
entry,
but
not
and
then
go
through
and
invert.
Then
the
data-
yes
I,
don't
know
what
folks
think
of
that
flow
or
if
that's.
A
C
Lose
great
I
mean
you
have
a
lot
of
content,
you'll
be
able
to
take
your
time
three
hours.
What's
gonna
go
fast,
but
yeah
lots
I!
Think
if
you
could
do
a
synthetic
I
would
probably
just
put
a
conductor
in
a
resistor.
You
know
I'm
good
set
rate
up,
and
then
you
will
see
the
cards
and
everything
cuz
they're
the
thing
with
layers
like
it's,
it's
it's
fine!
You
know
it's
like
it's
interesting
academically.
What
all
its
gonna
do
is
just
go.
C
A
One
question
I
have
so
right
now,
the
way
this
is
unfolding,
it
looks
like
I'm,
just
gonna
have
one
giant
notebook
for
a
three
hour
tutorial,
but
I
don't
actually
know
if
there's
a
clean
way
to
break
it
up.
The
folks
have
concerns
about
just
like
we
just
work
in
one
giant
notebook
or
do
you
think
it
would
be
helpful
if
I
try
and
break
so
yeah.
A
J
Especially
like
that
the
receiver
class
that
actually
stores,
if
he
meets
once
you
change
the
mesh
size,
then
like
your
house,
some
problems,
it's
a
yeah
I
think
there's
some
sort
of
like
just
like
a
and
the
coach
side.
There
could
be
potential
problems
and
you
may
not
want
to
because
I
would
say
if
you
have
a
small
bits
of
notebooks,
that
you
can
like
rerun,
usually
I
think
that's,
probably
easier.
Okay,.
G
Say
just
yeah
whatever
you
do,
I
just
think
it's
good
to
do
some
form
of
recap
and
and
show
where
you
are
in
in
the
presentation
before
you
move
to
the
next
next
thing.
So
as
long
as
you're
good
about
wrapping
up
this
chunk
before
moving
to
the
next
chunk,
so
that
people
know
where
they
are
in
the
span
of
this
three
hour,
presentation
I,
think
that
would
be
pretty
important.
J
Are
we
going
to
like
a
kind
of
do
that
comparison
with
the
like
a
previous
result?
Is
that
what
okay,
like
that'll,
be
sort
of
your
first
like
notebook,
then
like
I
kind
of
walked
through
all
that,
like
some
of
the
details
and
basically
run
like
without
maybe
explaining
too
much
and
see?
If
you
can
like.
H
A
J
A
Not
actually
sure
I
was
in
some
ways,
thinking
to
basically
get
to
this
point.
Where
now
we
can
forward
simulate
and
then
sort
of
take
a
detour
and
go
focus
on
the
synthetic
before
coming
back
to
the
field
inversion,
but
we
could
actually
basically
just
go
and
run
through
everything.
I
think.
J
A
J
H
C
A
J
Another
thing
that
you're,
probably
to
think
about
you
also
need
to
introduce
IP.
So
report
me
to
think
about,
like
that.
Are
you
going
to
choose
that
with
the
synthetic
like
kind
of
you
need
to
kind
of
choose
weird
when
that
comes
in
I?
Think
that,
in
terms
of
like
an
impact
that
I
deem
it
is
pretty
impactful
yeah
and
you've
got
pretty
good
story
behind
so
yeah
I'm,
not
at
the
mall
I'm,
not
100%,
sure.
A
That's
a
good
point:
do
you
think
there's
any
visual
that
I
can
show
for
IP
for
the
physics
to
get
folks
connected
with
on
I
mean
I,
look
back
at
what
we've
done
in
the
disk
and
such
but
I.
Don't
know
if
there's
anything
that
comes
to
mind.
That's
like
this
is
a
really
impactful
I'm.
A
J
Ago,
it's
actually,
you
can
like
it's
basically
reverse
currents
yeah.
So
the
physics
behind,
like
everything,
is
reversed.
So
it'll
reduce
like
effectively
reduce
the
currents
like
it
is
an
opposite
direction
of
like
your
primary
currents.
I
think
that's
a
double
win.
The
best
analogy
you
can
potentially
use,
but
I'm
yeah
I'm.
Not
sure
like.
Is
that
really
we're
close
to
go
in
that
much
detail
that
I'm?
J
J
I
Oh
right,
okay,
cuz
yeah
I
got
this
I
got
confused.
I
was
looking
at
the
EDI
loader
cuz.
We
were
having
actually
Devon
that
empty
script
is
awesome
like
that.
Mundy
code
is
pretty
cool,
but
we
are
having
a
issue
with
importing
it
now.
Just
looking
at
the
EDI
lower
our
loader
and
it
didn't
say
anything
about
the
Zed
component
up
or
down,
it
was
just
replacing
the
X
and
the
y.
So
I
was
like
oh
wait:
EDI
is
Zed
down,
so
I
wasn't
I
didn't
see
the
the
fix
there
so
yeah
thanks
that
clarified.
G
Yeah,
the
I
guess
I
I'm.
Writing
it
I
think
with
z+
down
pretty
pretty
sure.
I'd
have
to
go,
and
you
know
I'd
even
have
to
go
and
look
at
it,
but
yeah
I'd
have
to
look
at
it
again
right
right
now,
it's
we
haven't
really
leaned
one
way
or
another,
because
the
standard
seems
to
be
said
positive
down,
but
everything
in
simpang
we
really
want
to
have.
G
That's
the
industry
standard
and
that's
the
same
thing
that
that
I
think
jiff
tools
will
do
as
well.
And
if
you
look
at
the
derivations,
because
it's
still
in
a
right-handed
coordinate
system,
a
lot
of
the
derivations
will
will
work
for
both
so
I.
Think
if
you
are
just
looking
at
your
your
Z,
X
Y
components,
it
doesn't
matter,
but
I'd
have
to
double-check
that.
E
G
I
Because
the
apparent
chargeability
part
of
it
works
awesome,
it's
great,
it's
the
impedance.
It
seems
to
be
biasing
upper
upward
or
downward
when
we're
trying
to
run
the
inversion
on
some
but
yeah
god.
Yeah
I'll
just
play
around
with
these
components
here
and
see
stay
with
my
temper
inversion
when
I
was
saying
that
last
time
I
think
I,
just
like
one
of
the
components
is
off
so
yeah.
G
Well,
I
would
like
to
try
and
merge
the
1d
yeah,
the
1
DMT
stuff
into
the
branch
where
we're
working
on
3d,
MT,
stuff
and
but
I
guess
I'm
deciding
between
that
and
then
there's
also
trying
to
do
kind
of
a
stitch
to
1d
with
the
MT
code.
I
want
to
be
able
to
do
that
once
I
finish
with
the
e/m
stuff,
yeah.