►
From YouTube: 20220329 FPGA stand-up meeting
Description
FPGA Standup for 29 March 2022.
Plenty going on! Thank you to everyone for the very hard work in getting P4DX on the air.
A
So
we'll
start
with
paul,
tell
us
what's
going
on.
B
Hello,
nothing
all
haven't
done
anything
to
the
remote
lab
over
the
last
week.
I
think
it's
still
working
and
don't
have
any
blocks.
A
Oh
good,
okay,
so
I
think
that
that's
probably
good
news,
if
you
don't
have
much
to
report
from
remote
labs,
don't
have
to
reconfigure
anything
or
make
anything
work.
Then
I
think
that's
probably
good
news
is
that
you
were
feeling
as
well.
A
B
So
last
week
we
tried
to
debug
the
fpga
in
color
andre.
Well,
we
we
have
seen
that
there
there
are
some
difference
between
the
the
simulation
and
the
output
of
the
preto.
Then
we
can
see
that
on
the
simulation
compared
to
compared
to
the
new
radio
block
that
andre
use.
It's
not
exactly
the
same,
so
we
are
pointed
out.
B
B
B
So
it's
just
an
advance
when
the
output
of
the
encoder
will
be
okay,
then
we
have
a
good
filtering
and
it
could
be
a
easier
to
integrate
back
to
you.
A
C
Yeah,
so
if
there
is
so
he's
saying
tried
to
help,
but
you
know
he
did
most
of
the
work
so
yeah
there
is
the
difference
I
th.
So
I
think
the
difference
is
in
so
one
of
the
differences
right
is
the
like
the
block.
The
flow
diagram
has
the
header,
the
baseband
header,
the
it
uses
like
an
older
maneuver,
again,
radio
version,
but
the
yeah.
C
The
key
thing
is
the
header
that
messes
up
the
comparison,
so
I
I'm
working
on
make
the
making
the
environment,
like
the
the
repository
like
the
tests,
work
with
the
newer
nova
radio
version
and
removing
the
baseband
header.
So
ideally
I
think
they
do
match.
Like
I
remember
I,
I
got
to
a
point
where
they
matched
like
the
the
results
that
they
ever
found.
C
So
that
is
one
thing.
The
other
thing
is
it's
leonard
elder
leonard
right.
C
Yes,
I'm
terrible
with
names
but
aldi,
so
he
so
we
I
was
was
speaking
with
him
this
weekend,
so
he
gave
a
good
idea
to
like
so
I
tested
everest
c
function,
not
c
function
that,
like
the
program,
you
know
the
executable
that
we
used
to
send
data
and
they
basically
sent
a
ramp
sent
a
ramp,
and
I
looked
at
the
zla
and
you
know
engineers
is
correct
and
I
can
see
no
zero,
zero,
zero
one,
two
and
so
on,
which
means
we
can
send
a
custom
data.
C
C
So
the
last
thing
is
the
physical
air
framer.
So
you
say:
okay,
suppose
we
have
an
input
of
zeros
at
the
you
know,
beginning
of
the
flow
we
take,
the
physical
airframe
input
that
we
generated
generates
and
sends
that
on
pluto
and
see
if
that
works,
and
then
you
know,
if
it
doesn't,
then
it's
you
know.
The
physical
airframe
has
some
issue.
C
So
if,
if
it
does
work,
then
we
take
a
step
back
and
include
the
consolation
mapper
and
then
you
know
then
beat
interleaver
well
beat
interleaver,
you
know,
beating
deliver
doesn't
exist
in
qpsk
yeah,
essentially
move
backwards
using
a
new
rate.
I
I
think
that's
the
most
there's
probably
a
word
for
this.
Like
the
most,
I
don't
know.
C
C
Yeah,
so
I
I
want
so
I'm
focusing
on
getting
the
the
repository
in
a
state
that
you
know
we
can
get
to
the
divided
conquer
phase.
I
think
the
the
the
most
for
the
most
part
is
done
and
I
think
it
can
complete
in
the
following
days.
A
C
C
Like
a
similar
setup,
where
I
can
see
if
the,
if
I
receive
all
zeroes
in
the
decoder
you
know,
then
I
wouldn't
need
to
annoy
him,
but
I
I
I
don't
know
how
hard
it
it
it
is
to
get
that
in
the
lab.
A
No
yeah,
no
yeah.
I
think
I
I
remember
you.
You
mentioned
this
in
slack
that
all
zeros
is
you
know
your
assumption
about.
All
zeros
is
not
exactly
what
came
out
of
the
header
right,
yeah,
yeah
yeah.
I
think
this
is
this
is
an
under.
We
understand
the
the
problem.
This
is
sort
of
like
a
test
bench,
sort
of
and
yeah
yeah,
and
we
just
need
to
get
things
drived
up.
The
assumptions
are
not
quite
exactly
the
same
yet,
but
they
will
be.
C
A
And
and
now
are
oriented
and
it's
a
it's
a
really
good
good
set
of
progress.
I
think
it's
we're
on
the
right
track.
A
A
I
can't
tell
from
where
I'm
at.
If
I
can,
somebody
nod
or
give
me
a
thumbs
up.
A
Oh
good,
okay,
so
it
should
say:
fpga
stand
up
29
march
blah
blah.
Is
that
what
you're
saying
yep?
Oh
good,
okay,
so,
okay?
So
we
are
our
matlab
startup,
the
matlab
or
mathworks,
is
the
company
and
we
applied
for
and
we
were
able
to
get
into
their
so-called
startup
program,
which
is
fantastic.
This
is
a
really
great
thing
that
they
do
now
they're
aimed
at
more
commercial
startups.
A
So,
as
we've
kind
of
explained
before
there
isn't
really
a
matlab
license
for
non-profits
like
us,
the
home
license,
which
we've
been
using
up
to
this
point
specifically
excludes
research,
which
is
what
we
do,
but
it's
a
great
thing.
So
if
you
don't
have
a
home
license,
it's
only
150
for
matlab
and
simulink
and
you
get
a
wide
variety,
not
all,
but
you
can
have
access
to
a
heavily
discounted,
wide
variety
of
toolboxes.
A
A
Lots
of
people
helped
convince
matlab
that
they
should
take
a
chance
on
us
and
do
a
sort
of
a
pilot
with
a
non-profit,
so
we've
got
another
11
months
or
so
to
make
the
fullest
use
possible
this
program
and
we're
focusing
on
hdl,
coder
or
hardware
description,
language
coder,
it's
a
toolbox.
What
it
does
is
it
takes
matlab
code,
it
converts
it
to
hdl
and
it
handles
conversion
to
fixed
point
and
then
performs
a
very
verification,
validation
and
simulation
and
synthesis.
A
But
it
relies
on
tools
that
you
already
have
so
the
simulation
synthesis
tools
are
things
like
model,
sim,
vivado
or
cadence,
or
anything
like
that.
So
it
it
doesn't,
do
it
in-house,
but
sets
up
and
uses
and
leverages
the
vivado
for
us.
It
can
also
program
a
hardware
target
using
vivato
and
it
could
do
fpga
in
the
loop
along
with
two
other
types
of
test.
A
There's
an
example:
here's
the
here's,
the
link
for
the
example
that
we've
started
out
with
and
here's
the
notes.
So
this
is
the
hdl
code
or
hello
world.
It's
code
generation
of
of
a
symmetric,
finite
imp,
infinite
response
filter.
So
it
sets
up
a
filter
and
then
what
you
do
is
you
generate
that
that
filter
is
hdl?
So
I
I
read
the
through
the
code
and
here's
the
filter.
It's
actually
doing
so
you
can
see.
A
It
really
is
a
symmetric
fir
filter,
that's
that's
the
one
and
what
it
does
is
you
set
up
the
algorithm
and
you
set
up
a
test
bench
in
with
two
matlab
scripts.
So
these
are
two
separate
matlab
scripts,
dot
m
files
and
that's
how
you
feed
the
hdl
coder
toolbox.
A
Here's
notes
about
how
you
configure
what's
called
a
workflow
advisor,
which
is
a
a
really
neat
and
so
far
as
I
can
tell
well
done,
workflow
assistant
that
walks
you,
through
the
process
of
setting
up
your
your
matlab
code
in
order
to
do
things
like
synthesize,
simulate,
verify,
validate
and
and
then
even
clean
up.
So
it's
all
done
in
temporary
folders.
It
doesn't
mess
up
your
matlab
script.
A
A
It
is
a
little
bit
weird
to
see
vivado
in
a
matlab
command
window,
but
that's
what
you
end
up
getting
so
matlab
is
essentially
commanding
vivato
to
do
stuff
for
it,
and
you
can
see
in
the
upper
window,
essentially
the
the
sort
of
the
upper
center
window.
You
can
actually
see
the
generated
hdl
and
it
is
recognizable
relatively
human,
readable
hdl.
I
was
really
kind
of
stunned
at
how
complicated
this
fire
filter.
A
It
was
in
hcl
there's
an
awful
lot
of
you
know,
manipulation
and
handling
of
the
the
filter
coefficients
for
this
filter
and
thinking
back
like
in
my
professional
career.
I've
never
really
had
to
handle
this
sort
of
stuff
before
it
was
always
dsp
that
was
already.
A
You
know,
made
lots
of
not
a
whole
lot
of
filter
coefficients
other
people
on
the
team
did
that
so
so
I
was
like
wow.
Okay,
no
wonder
they
had
such
a
hard
time
at
work,
so
my
empathy
level
has
gone
up
for
for
dealing
with
this
and
when
I've
done
iq
gain
phase
correction.
Algorithms.
In
the
past,
I've
had
to
deal
with
this
where,
where
you
have
to
do,
you
know
carefully,
manage
your
fixed
point
null
so
anyway,
it
did
it
it
it.
It
came
across
it
got
tested.
A
A
So
this
this
toolbox
forces
you
to
write
a
test
bench
from
the
get-go,
that's
how
it
works,
and
it's
pretty
clear
from
the
example
in
the
documentation
that
you
can't
just
simply
write
some
ip
and
have
it
turned
into
hdl.
Unless
you
want
to
go
by
hand
and
do
a
whole
bunch
of
stuff
in
the
configuration
that
a
test
bench
would
provide
for
you.
A
A
The
hardware
server
right
now
is
connected
to
the
pluto,
and
even
though
I
started
up
another
hardware
server,
it
doesn't
really
know
it's
not
smart
enough
to
just
kind
of
have
esp
and
figure
out
what
I
want
as
a
programmer.
So
that's
where
it
stopped,
but
you
can
see
it
made
it
all
the
way
to
the
bitstream.
A
It
started
attempting
to
load
the
bit
stream
onto
the
hardware
target
that
I
had
specified
as
the
zcu
106
chip.
So
the
problem
boiled
down
to,
I
think
the
pluto
is
still
is
connected
up
to
karapi
and
as
well.
It
should
be
because
that's
where
we're
getting
lots
of
work
done-
and
it's
just
now
a
question
of
like
figuring
out
how
to
use
the
hardware
server
or
maybe
moving
all
of
this
to
to
a
third
virtual
machine.
A
You
know,
so
that's
just
my
fault.
I
I
haven't
ever
had
to
deal
with
two
targets
connected
up
to
the
same
hardware
server
or
the
same
version
of
avato
before
I
expect
it
to
solve
problem
I'll,
go,
educate
myself
and
there's
also
a
lot
of
reports
generated
from
hdl
coder.
That's
really
pretty
useful.
I
say
that
the
process
was
painless,
but
I
was
expecting
it
to
be
a
case
of
you
have
to
construct
your
matlab
code
correctly
for
this
toolbox
to
work.
A
It
was
relieved
to
hear
that
we
understood
that
it
wasn't
magic.
This
is
a
very
complex
tool
that
gives
you
a
tremendous
amount
of
traction
on
hdl
that
allows
people
that
don't
know
verilog
or
vhdl
to
participate
in
fpga
and
asic
design,
but
you
have
to
set
up
your
matlab
code
correctly.
So
in
other
words,
we
already
knew
this,
but
we've
just
gotten
a
tremendous
amount
of
confirmation,
not
just
from
humans
at
mathworks
over
on
the
phone,
but
also
from
walking
through
one
of
the
more
complicated
examples
that
they
have.
A
The
the
toolbox
comes
with
a
suite
of
examples,
and
you
know,
I
think,
probably
working
through
a
few
more
of
them
before
we
attempt
to
turn
the
m17
code,
because
that's
the
first
real
target
that
we're
going
to
do
is
take
the
m17,
which
is
our
uplink
protocol,
to
turn
that
into
hdl
and
we'll
be
able
to
compare
it
against
some
human
written
hdl
code
on
the
pink
fpga
board
from
wx90o
who's.
Who's
done
a
tremendous
amount
of
work
here.
A
D
Yeah
sure
here
we
are
this
great
interest
to
me,
because
you're
hitting
on
two
subjects
that
are
giving
me
a
great
deal
of
trouble
right
now.
I
have
a
project
that
I'm
trying
to
do,
and
this
is
a
mock-up
of
the
project,
a
pluto
with
an
antenna
and
a
bias
t
and
a
battery
pack
on
the
back,
and
it
receives
signals
from
satellites
sitting
here
in
the
base.
D
With
most
it
has
its
own
fpga
code
in
it,
not
the
pluto
code
that
came
with
it,
and
I
have
a
couple
of
these
and
the
two
one
project
that
I
was
supposed
to
do
was
to
take
the
pluto
board
and
modify
it,
that
is
to
say,
design
a
new
pluto
board
to
put
in
the
pluto
case.
That
has
everything.
A
pluto
board
has
some
other
circuitry
relating
to
the
battery
pack
charging
and
regulation
and
put
it
all
inside
a
pluto
case
and
have
a
pluto
on
steroids
that
can
at
least
receive
satellite
signals.
D
That's
what
this
one's
designed
to
do,
but
also
possibly
transmit
them
on
a
handheld
device
with
eight
hours
of
battery
capability,
and
that
was
a
great
project.
I
built
that
you
can't
buy
the
sync
processors
that
are
inside
the
pluto
or
any
sink
processors
anywhere
in
that
whole
suite
of
processors,
70
20,
30
40,
whatever
they
are,
you
can't
buy
any
of
them
they're,
not
in
production,
so
we
called
what
was
then
xilinx.
D
I
noticed
my
stock
portfolio.
Xylanx
has
disappeared,
they've
been
purchased
by
amd
to
the
future
of
silence,
but
in
any
case
fold
them
up,
and
they
said,
oh
well,
why
don't
you
just
go
to
this
other
series
and
everything
will
be
fine,
and
we
said,
oh,
my
dear
lord,
we're
going
from
20
30
parts
to
100
to
a
thousand
dollar
parts,
and
this
whole
product
is
not
going
to
make
sense
at
that
point.
So
that's
one
thing
that
I
have
in
common
with
what
you're
doing
another
thing
is
math.
D
I
did
not
renew
my
subscription
to
simulate
this
year,
although
I
still
could-
and
I
renewed
matlab-
because
I'm
still
using
it,
but
what
I'm
doing
now
is
I've
replaced.
D
I
I'm
rewriting
some
of
the
matlab
m
scripts
that
I've
used
gps,
24,
7
workbench
here,
where
I
have
a
lot
of
equipment
set
up
and
I
collect
the
data
and
produce
pdf
reports
with
matlab
very
nice.
D
Reports
and
I've
made
very
good
progress
now
on
the
world
of
machine
learning,
as
as
implemented
by
open
source
modules
like
panda
and
python
and
numpy
and
psi
learn
and
map.
D
You
know
math
print
pot,
whatever
it's
called
and
and
I've
done
a
pretty
good
job
of
that
unfortunate
whole
world,
which
is
new
to
me
and
I'm
very
excited
about
it,
because
I
can
do
a
lot
of
things
pretty
much
almost
everything
I
can
do
in
matlab.
I
can
do
there,
I'm
finding
the
open
source
world
to
be
a
little
bit
tricky.
D
So,
when
I
go
in
and
and
looked
at
python
and
then
the
syntax
of
these
other
modules
to
do
the
same
thing,
the
syntax
changes,
it's
probably
typical,
of
the
open
source
world,
where
all
these
modules
produced
by
all
these
different
people
do
complementary
things
but
they're
enough
different
to
be
confusing.
Sometimes,
nonetheless,
it's
worth
it.
I
think,
to
go
to
the
open
source
world
and
get
my
future
data
analysis
in
that
world.
D
D
A
Yeah
they
are
shockingly
expensive
for
those
of
you
that
don't
know
the
the
quote
for
hdl,
coder
and
and
the
things
that
are,
and
as
I
think
that
that
rick
will
probably
agree
with
me.
It
isn't
always
it
is
it's
never
just
one
module
right,
it's
always
that
module,
and
then
you
need
about
three
others,
because
there's
dependencies
there's
an
awful
lot
of
them.
There's
nearly
a
hundred
of
these
toolboxes.
A
So
the
thing
that
we
were
most
interested
in
in
trying
out
and
using
actually
requires
three
or
four
others.
So
when
you,
when
you
get
them
all
together
and
you
go
get
a
quote
to
just
wander
in-
and
you
say:
okay,
we
want
to
do
this
say
commercially.
Then
it
ends
up
being
nearly
thirteen
thousand
dollars
and
that
that
is
actually
a
low
end.
A
That's
that's
it's
thirteen
thousand
dollars,
in
addition
to
do
you
buying
it
as
a
commercial
customer,
so
you
can
see
how
the
home
license,
which
is
an
effort
that
I
was
involved
in
trying
to
make
happen.
You
know
for
150
bucks,
you
can
get
the
matlab
license,
but
they
won't
give
you
hdl
coder
the
reason
that
they
don't
give
you
hdl
coders
because
of
the
amount
of
customer
support
that
has
been
required
because
it's
complicated
so
those
of
us
that
are
that
are
in
open
source
that
also
are
in
fpgas.
A
We
get
kind
of
tossed
out
with
the
bathwater,
because
we're
such
a
small
community
really
so,
hopefully,
with
this
particular
experience,
we
may
be
able
to
make
a
a
case
that
things
like
hdl
coders
should
be
40
bucks
for
people
that
are
not
going
to
use
it
commercially,
that
are
not
going
to
use
it
in
an
academic
sense
that
are
not
going
to
use
it
for
r
d.
You
know:
okay,
fine
at
least
that's
a
step
forward,
because
those
are
the
people
that
are
backbone
of
open
source
work
and
yes
it.
A
You
know
it
gives
you.
You
know
just
a
a
boost
right,
so
so
it
allows
you
some
some
tremendous
leverage
from
matlab
to
hdl
coder
that
but
like
for
people
like
you
or
me,
or
ever
easter,
swato
or
any
of
the
people
working
on
this
can,
if
they,
if
you
can
write
in
hdl,
natively
and
or
v,
you
know
hdl
of
any
type,
vhdl
or
verilog.
That's
still
going
to
be
the
best
way
forward.
A
The
code
that's
produced
is
is
actually
readable
and
good,
but
I
think
we'll
see
over
time
whether
or
not
it's
as
good
as
the
code
that's
being
produced.
So
part
of
part
of
the
reason
for
doing
this
is
to
see
okay,
take
this
proprietary
tool,
use
it
and
then
compare
it
to
an
open
source
implementation
of,
say,
m17,
uplink
protocol
and
like
okay,
let's
see,
can
you
do
as
good
with
it
and
if
you
can
do
is
good,
then
hey
octave,
the
open
source
matlab.
A
A
We
are
proceeding
and
marching
towards
that
goal,
but
we
are
also
looking
at
improving
open
source
tool
flow
and
the
only
way
that
we
can
really
get
to
know
where
we
need
to
be
is
by
you
know,
bravely
some
would
say
someone
stupidly
blundering
into
things
like
this
and
and
just
using
the
stuff
that
is
used
in
industry
and
going
and
seeing
what
we
can
achieve
with
it,
so
that
we
have
some
experience
with
what
is
what's
on
the
proprietary
side,
because
that
is
a
standard
that
open
source
tools
should
aspire
to
deliver,
and
you
brought
up
something
very,
very
important.
A
There
is
a
huge
inconsistency
in
not
just
user
experience,
but
also
user
interface
and
configuration
and
quality
and
completeness
across
open
source.
So
the
tools
for
for
machine
learning
are
exceptionally
good
and
open
source.
It's
an
excellent
place
to
to
leverage
you're,
not
wrong
there,
but
yeah.
You
will
see
some
inconsistency,
because
each
there's
there's
there's
not
there's
different
enforcement.
A
So
to
say
so
to
speak
of
you
know,
there's
not
the
same
desire
for
consistency
or
need
for
it
from
the
open
source.
Community.
Being
all
you
know
a
lot
of
cases,
its
grassroots
are
built
from
the
bottom
up.
So
thank
you.
So
much
for
sharing
and
and
talking
we've
raised
a
number
of
of
intersecting
issues
here
so
back
to
you
for
a
final.
D
No
one
other
comment
I
have
is
that
the
worst
doing
fpga
work
are
software
people,
the
majority
of
software
engineers.
I
know
don't
understand
that
this
is
not
a
microprocessor.
D
The
some
of
the
best
programmers
are
people
with
primarily
hardware
expertise
who
can
understand
that
this
isn't
a
sequential
machine
with
some
interrupts
and
things
this.
D
This
is
really
a
hardware
matrix
that
you
can
do
hardware
things
with,
and
yes,
so
if
you
can
design
that
fir
iir
filter
in
hardware
with
with
old
lsttl
chips,
then
you
can
probably
program
at
fpga
really
really
well,
but
if
you
think
you're
going
to
do
it
with
c
code
and
yes,
the
modules
translate
c
code
into
fpga
you're
not
going
to
get
a
particularly
good
answer,
and
I'm
not
sure
I
don't
know
anything
about
matlab's
transition
of
matt
m
code
into
fpga
code,
but
m
code
looks
an
awful
lot
like
straight
programming
to
me.
D
Both
we
had
a
really
new
and
understood
hdf
how
to
program
hdl
correctly,
but
that's
harder
to
do
than
you
might
think.
A
No,
it
does
things
worth
doing
are
rarely
easy.
They
require
an
awful
lot
of
seat
time
and
that's
going
to
continue
to
be
true,
no
matter
how
hard
we
try
to
make
it
easy
and
accessible,
and
you
know
that's-
we
just
have
to
do
our
best,
but
so
far
the
hdl
coder
has
exceeded
my
expectations,
which
are
not
very
high.
It
really
is.
I
was
expecting
it
to
be
more
like
hls,
the
so
high
level
synthesis
from
vivado.
A
A
This
seems
to
be
a
better
match
for
this
type
of
conversion
than
than
just
c,
which,
yes,
you
can
write
c
code
and
make
it
work
well
in
hls.
I've
seen
it
done
before,
but
it's
it
requires
a.
I
think
more
effort
on
the
part
of
the
programmer
so
we'll
see
and
the
the
task
here
is
to
document
it
fully
and
and
to
you
know,
to
to
push
forward
and
to
improve
open
source
tool
flow,
because,
if
they're
doing
this
in
matlab,
this
could
be
done
open
source.
A
And
you
know
we
see
a
growing
fpga
open
source
community
and
they
have
a
lot
of
impediments
and
roadblocks
and
difficulty
challenges.
So
any
sort
of
overall
kind
of
like
a
tool,
flow
innovation
or
an
awareness
of
what's
what's
possible.
A
And
you
know
if,
if
we
see
some
design
patterns,
that
really
should
be
honored,
then
yeah
we'll
we'll
do
our
best
to
kind
of
raise
that
up
as
a
as
something
for
the
open
source
community
to
to
tackle.
So
that's
part
of
the
the
mission
at
ori.
It's
part
of
what
we're
doing
in
the
meantime,
though,
if
we
can
use
this
tool
to
make
some
really
good
hdl
that
helps
us
with
our
uplink,
then
right
on
we'll
keep
doing.
D
A
That's
a
shame.
Well
I
mean,
if
you
want
to
work
on
open
source,
amateur
radio,
satellite
stuff,
then
the
remote
labs
is
there
for
you
to
use
so.
A
Very
good,
hey,
very
good.
That's
all!
We
need
to
hear
all
right
cool
all
right,
any
other
comments,
questions
or
any
resources
that
anybody
needs
or
any
roadblocks
that
anybody
has
before
we
close
the
meeting
for
today.
All
right,
thank
you,
everybody
we'll
keep
it
up
and
we'll
be
back
here
next
week.
Please
check
in
on
slack
to
see
all
the
really
cool
stuff
the
bit
the
the
mismatch
that
was
discussed
earlier
between
every
and
swato.
There's
a
really
good
image
of
the
problem.
You
can
see
that
everything
works
great.