CSI Computer System/Remote Interface Unit Acceptance Test Results
/
/
NASA
Technical
Memorandum
104218
_
CSI COMPUTER
SYSTEM/REMOTE
INTERFACE
UNIT ACCEPTANCE
RESULTS
('_ASA-TM-IO431d)
I'vIrFRf;ACE
UNIT
(t,_A_A)
17_
i-,
CSI
CqMPUTE_,
ACCEPTANCtI'E_T
TEST
SYST_M/RFMQTE
RESULTS
CSCL
20K
Nq2-24)4_,
G .3/..,9
Dean
W. Sparks,
March
1992
National Aeronautics
Space Administration
Jr.
and
Langley Research Center
Hampton, Virginia 23665-5225
Acronyms
The following
acronyms
are used throughout
AID
Analysis
A/D
this document.
and Computation
Division
- Analog to Digital
AP
Array
Processor
CAMAC
-
Computer
CSI
-
Control/Su'uctures
CCS
-
CSI Computer
CDPP
-
Console
CEM
-
CSI Evolutionary
COFS
-
Control
of Flexible
Structures
CPU
-
Central
Processing
Unit
Automated
System
Debugger/Prom
IX_
Direct Current
Digital
Signal
Programmer
Model
Digital to Analog
-
and Control
Interaction
D/A
DSP
Measurement
Processor
EDS
Excitation
FIR
Finite Impulse
Response
FSGB
Flight
Software
and Graphics
GSE
Ground
Support
Equipment
GSET
Ground
Support
Equipment
HRM
High Rate Multiplexer
UR
Inifinte
po
Input-Output
RIU
Remote
Interface
SED
System
Engineering
SSRL
Space Structures
and Damping
Impulse
Subsystem
Computer
Branch
Terminal
Response
Unit
Division
Research
Laboratory
Purpose
The purpose
(CSI)
Computer
Structures
of this document
System
Research
extension
(CCS)/Remote
Laboratory
the CSI Evolutionary
of the earlier
is to report
Model
Interface
(SSRL)
(CEM).
on the testing
of the Control/Structures
Unit (RIU),
which
was installed
for use in conducting
real time control
The test work performed
on the CCS/RIU
test work performed
on a duplicate
Interaction
in the Space
experiments
on the
in the SSRL is an
CCS and the RIU in the Flight Software
and Graphics
Branch (FSGB)
Laboratory
at the Analysis
March
1991.
This document
is organized
in the following
overall
CCS/RIU
system
is given.
Then,
the various
and Computation
manner.
software/control
computations,
open loop, closed
standalone
- are discussed.
Next,
outlining
solutions
and recommendations
closes
the document.
System
Description
The CCS is intended
system
with spaceflight
a section
for improvement
to allow
researchers
qualified
components.
to conduct
for the cancelled
Control
consists
of four major
components:
the Console
the Ground
Support
Interface
System
Equipment
Computer
Unit (RIU) 2 and a dumb
function(s)
Console
of each component
is briefly
Debugeer/Prom
Proerammer
XT and a prom burning expansion
the EDS via a RS 422 serial interface,
the mass memory
and MIL-STD
necessary.
Another
of quickly
halting the control
Excitation
performing
function
and
System
the real time control
up to 100 states.
Sensor signals
respectively,
VAC
contained
wall socket
section
tests on a computer
by SCI of Huntsville,
flight project,
Programmer
(CDPP),
Equipment
In addition,
two other components,
have been interfaced
- The CDPP
Processing
is to provide
the
(GSE),
and
the Remote
with the CCS.
consists
The
of two parts: a PC
used to download
Unit (CPU)
the user access
and thus the control
rEDS)
This
(AP) boards,
1750A
when
test, in case of emergencies.
mass memory,
MIL-STD
all of which
are spaceflight
qualified.
and is capable
of handling
controllers
axe received
and
also has been qualified
in the GSE (the other CCS components
2
of the EDS,
to the EDS for the pro'pose
draw their power
with
for space operations.
(112 W) from one of three direct
sources).
to
the proms on
CPU,
actuator commands
on 1553B data bus, which
boards
software
unit is the 'brain' of the CCS,
It houses
law computations,
its power
and
below:
and the computed
The EDS is unique in that it derives
supplies
Central
computations.
control
a short summary
the prom burner is used to re-program
and array processor
the matrix
their
(COFS)
The PC XT is primarily
law computations,
The AP performs
transmitted,
1750A
- hardware
problems,
Support
(CPDD)
while
of the CDPP
Damnin_
1553B communication,
explained
of the
and RILl
the Ground
terminal,
chassis.
Structures
Debugger/Prom
(GSET).
graphics
Finally,
in
filtering
development
real time control
of Flexible
(EDS)1,
Terminal
on the CCS/RIU
The CCS was manufactured
originally
and Damping
the system's
(ACD)
a brief description
loop, RIU digital
is presented.
Alabama,
Excitation
First,
tests performed
acceptance,
Division
era'rent (DC) power
from standard
120
Ground
Sup m)rt Equipment
input-output
power
(I/O) interface
supplies,
1750A
between
CPUs
IEEE 488 communication
The initial power
however,
source
supply
bypass
the I/O interfaces
purpose
of testing
GSET.
Thus,
in the SSRL,
Support
that the I/O functions
monitor,
parameters
which
selections,
sampling
excitation
types
communicates
and serves
define
and sources,
control
biases.
commands)
off of the 1553B
Finally,
there
to transfer
the data to other
(25 megabytes
Remote
systems
at present)
Interface
Unit
Built in-house,
input channels,
multiplexed
analog
into two 12-bit
low pass filtering,
filtering.
With
the DSP,
called
"standalone"
by the System
12-bit
and
Division
and actuator
disturbance
to the user and computes
pre-
signals
converts
and actuator
it into ASCII
interface
has been
has enough
(SED),
format.
installed
free hard disk
and filtering
duties
this unit provides
16
as well as 8 output
channels,
each).
The RIU also provides
limited
and a Digital
Signal
Processor
for digital
to perform
i.e., with the CCS out of the loop.
the necessary
A 28 VDC,
(DSP)
real time control
board
tests itself
in a so
In the normal
mode,
the RIU
7 amp power
supply,
also built by
primarily
for viewing
power.
- A dumb
Via a standard
i.e., sensor
in the
(4 channels
SED, provides
the test data.
the user enters
the data acquisition
A/D converter,
with the CCS via the 1553B bus.
Terminal
PC with
also
an ETHERNET
Engineering
communicates
Granhics
messages
- The RIU performs
the RIU has the capability
mode,
based
The GSET
test runs.
for anti-aliasing,
the EDS
scale factors.
to hold data from multiple
D/A converters
powers
filter selections,
The GSET
to a single
directly
the real time test data (sensor
for post-test
to
by the
which
analysis.
(RIU)
for the CCS.
multiplexed
and instrument
on the GSET;
was made
more efficiently
to be conducted,
law matrices
software
wall
It was felt, for the
Here,
data bus onto its hard disk drive,
is no analysis
a decision
to the CCS.
and digital
records
120 VAC
is a 80286
error and warning
and
for test data storage.
- The GSET
experiment
1553B
test operations.
of test, analog
the GSET
Currently,
space
as the user interface
and software
test sensor
normal
three DC
and 30 amps,
in the SSRL.
supply
eGSET)
the real time control
rate and length
hardware
the one power
Terminal
phase,
could be handled
will be used during
Equipment
system
that a standard
the CCS development
by the GSE for operations
configuration,
for I/O processing,
tape drive
revealed
a Shuttle
It contains
for the GSE called for 240 VAC
During
provided
built to simulate
test article.
(HRM)
magnetic
with the manufacturer
was sufficient.
in the current
a 20 inch color
and a 9-track
specifications
is the only part of the GSE which
Ground
the CCS and the experimental
interfaces,
discussions
socket
- The GSE was originally
and a High Rate Multiplexer
requirement
subsequent
fGSE)
ROLM
graphics
terminal
phone/data
has been provided,
link, the user can log on to other
3
systems,
to which
the test data has been previously
sent (e.g., a VAX with PRO-MATLAB)
for post-rest
analysis
work.
Figure
1 shows
Figure
2 is a schematic
follows.
Analog
signal
conditioning
the major components
sensor
showing
signals
to the RIU, where
they are convened
computations.
The resulting
1553B,
they are convened
the Control
CCS
Trailer
and finally
the Control
to analog
Test
SCI has provided
a detailed
procedure
The full hardware
acceptance
in August
1990, and a portion
specifically
the EDS,
was still functioning
performed.
operations,
Only those
The hardware
the SSRL,
signals.
other
through
the safety PC and
as required.
These
signals
via the 1553B bus for use in control
are transmitted
These
are
law
to the RIU down the
analog
actuator
commands
purpose
of testing
the functionality
properly.
supply
and power
-
array
processor
-
1750A
CPU
-
1553 bus communication
-
shared
are sent into
A and B contain
the general
acceptance
tests.
The general
tests, while
the error log file contains
the current
for the EDS,
system
of
delivery
into
that the CCS hardware,
CCS will not use the GSE
the GSE I/O tests were not
and they covered
the following:
supply
(AP)
between
tests indicated
properly
Since
to ensure
the EDS were conducted,
voltage
memory
after initial
of this test was repeated
-
were functioning
The data flow is as
are then sent from the main patch panel
test was performed
than a power
tests involving
acceptance
Trailer,
and filtered
for the express
the SSRL
actual
signals
the SSRL.
out to the CEM.
Acceptance
during
within
signals
Trailer
connections.
Control
actuator commands
Hardware,
CCS hardware.
sensor
to digital
digital
system
into the SSRL
The analog
then sent to the CCS back inside
where
the CCS/RIU
come
amplifiers.
and their respective
the AP and 1750A
that the EDS, as well as the other CCS components
and were ready
for further
log and error log file printouts,
log file reflects
a more
CPU
testing
with the RIU.
respectively,
and on-screen
detailed
of test result messages.
4
Appendices
for these CCS hardware
the user inputs
listing
in
responses
from the
Dummy
Control
Computation
After the completion
of the CCS hardware tests, a series of so called
experiments
were executed
experiments
because
computational
matrices
Tests
on the CCS.
they were designed
software,
little thought
was placed
computed
actuator commands
recorded
sensor
on exercising
signals
laboratory
control
- since the computed
was given
emphasis
tests, nine in all, are referred
to test the CCS software,
and not as actual
were made up arbitrarily
to the CEM,
These
or stability
All the control
were never
law
transmitted
laws.
Instead,
of the CCS software.
The sensor
signals
and CCS
for each test for post-test
analysis.
to a VAX workstation
"dummy"
the EDS
of these control
the features
was transferred
experiments.
control
to as "dummy"
specifically
actuator commands
to the performance
were recorded
"dummy"
Each
set of
for use in a PRO-MATLAB
simulation
of the corresponding
control experiment,
with the CCS computed
commands
being
compared
to the actuator commands
obtained
1 shows
the major
parameters
each of the "dummy"
in the simulation
actuator
for verification
purposes.
Table
entries
are as follows:
of sensors
column
column
and actuators
1 contains
feedback;
excitations
which
Table
Test
No.
the '!dummy"
used in the test, respectively;
5, the type of control
types of disturbance
that describe
of
column
of
column
columns
2 and 3, the number
4, the CCS digital
6, the number
of controller
The column
sampling
rate;
column
7, the
states;
were used.
1. Dummy
No.
test number;
tests.
computation
Sampling
No.
control
tests.
Control
Control
Y.V.nt
Excitation
States
1
4
8
150
output
O
p
2
1
1
150
state
2
none
3
1
8
100
output
0
s,p,r
4
8
8
100
state
100
s,p,r
5
8
8
100
state
2
p
6
8
1
150
output
0
s
7
8
1
150
state
50
none
8
8
6
150
state *
100
s,r
9
8
8
150
0
none
p - pulse;
s - sine; r - random
output
**
*
-
The controller
involved
**
-
was not turned on during
the defined
The control
gain matrix was set equal to the identity
"dummy"
control
first time, a signal
sensor
being
split into the appropriate
Figure
were actually
was connected
mimic
with this particular
A single
constant
configuration,
directly
frequency
number
of sensor
generator,
the EDS computed
MATLAB
simulated
actuator
verifying
themselves
the signals'
executed
i.e., the actuator
twice
commands
output
commands
from the signal
for each test.
used to
generator
was used,
It should
the SSRL
matched
by plotting
and amplitudes
the known
this
that
Trailer (see
In all nine tests with
their counterpart
their time histories
against
be noted
Control
the 1553B data bus.
commands
For the
were
for the same set of re.corded sensor
were checked
frequencies
through
actuator
on the CCS/RIU.
to the RIU, and sine waves
channels
2), rather, the data flow was exclusively
signals
matrix,
the data flow did not go through
the signal
sensor
only
signals.
experiments
generator
signals.
commands
excitations.
were equal to the sensor
These
the test, thus, the actuator
PRO-
signals.
The recorded
on the VAX
parameters
and visually
set on the signal
generator.
The next round of tests was conducted
connected
to the RIU.
The CEM
CCS/RIU,
_were thrusters
#1-8,
pair of thrusters
document.
The CEM
were recorded
histories
act in unison,
with the CEM
actuators
referring
to the eight
excited,
thruster
PRO-MATLAB
command
simulation,
time histories,
commanded
(duration
controller.
matched
The CCS actuator
for each thruster.
This indicated
performing
experiments,
command
This result
that the CCS software,
as desired,
As a further
was
study
data from the aocelerometers
Figures
3a-d
show the time
for test #1, while
Figures
4a-p
show
i.e., 0.0067
was true for the other
6
turned
on
4a) reflects
a
at the initiation
of the
simulation
actuator
command
"dummy"
control
experiments.
eight
software,
was
could commence.
of the CCS,
but varying
#1 (Figure
seconds)
the EDS computational
capability
and 8 outputs,
with the controller
for thruster
and the PRO-MATLAB
into the computational
the
by the CCS in real time and by thc
and that open loop tests with the CEM
each with 8 inputs
of this
for the remainder
tests.
at 20 seconds
particularly
to the
actuators
control
as computed
1 time sample,
connected
each
Test #1 ran for 40 seconds,
Note that the 1 lb spike
#1-8,
pairs on the CEM 3. Since
and the free response
for the last 20 seconds.
1 lb pulse
not yet physically
to as single
respectively,
respectively.
servo accelerometers
air thruster
and used in real time in the nine "dummy"
of the first four accelerometers,
corresponding
for the tests, though
they will be referred
was manually
sensors,
control
another
matrices'
set of "dummy"
sizes,
control
were executed
todetermineachievablethroughputspeedson the CCS.
particular
Open
C contains
Loop
the CCS/RIU
transmitted
Tests
with
software
the
computational
software
from
testing.
the CCS to the CEM,
excitation
commands
of commanding
are as follows:
constant
frequency
random
excitation.
duration),
and uniform
to output
any of the above
commanded
to perform
Along
with the digital
the GSET
commands
precautionary
measure,
For these
law computations
were used to drive
the CEM.
types of excitations
sine waves,
single
pulses
In the CCS software,
three excitations
during
per test.
accelerometer
and thruster
recordings
were also available
to the CEM,
verify
Only
after these open loop commands
All three types
of excitation
in the following
disturbing
turningoff
were tested
paragraphs.
the CEM
the excitations
and allowing
seconds
of the test.
this test.
Note the open loop responses
time histories
Figures
at 7 seconds,
command
time histories
follows:
thruster
0.9 lbs, frequency
and thruster
5a-h show
indicating
#3 - amplitude
#7 - amplitude
on the CEM,
and sample
for the first seven
the CEM's
motions
growing
6a-d.
The thruster
of 2 lbs, frequency
of 11.9381
thruster
of 0.5 lbs, frequency
7
of 0.9111
for the 3 types
are
The test
decay
for the remaining
rad/s;
#1-8,
off.
of 2 lbs, frequency
23
for
in several
The four thruster
commands
thruster
then
respectively,
and the spikes
sine wave
rad/s.
on the
of the experiment,
were commanded
#6 - amplitude
In fact, as a
seconds
from 0 to 7 seconds,
in Figures
signals
from the CCS
excitations.
of accelerometers
the thrusters
rad/s;
accelerometer
on
on the strip charts to
results
to freely
the time histories
recorded
were confirmed
was sine wave
when
are shown
of 10.6814
tests.
with four thrusters
which were
recorded
on subsequent
The first type tested
can only be
commands
Trailer
were they sent out to the CEM
period
on the open loop tests.
but merely
CCS
can be commanded
analog
were not sent out of the Control
strip charts,
The current
each thruster
for the first few open loop tests, the thruster
the open loop commands.
were
were performed,
any actuator
commands,
checks
to
for use as disturbances.
of the corresponding
as further
were connected
(of one sample
a test, although
one type of excitation
signals
the thrusters
tests, thruster commands
no control
three different
from each test, strip chart
and thruster
tests were passed,
however,
is capable
involved
on these
CEM
for open loop command
only the programmed
given
the details
tests.
Once the CCS control
They
Appendix
were as
#4 - amplitude
of 0.9111
of
rad/s;
The next excitation
above
test used single
for the sine wave
plot in Figure
7, were
excitations
respectively.
Figures
commanded
and thruster
to the pulses
outputs
uniform
random
variable
(between
with the variance
the zk's, while
the product
excitation.
command
of accelerometers
Thrusters
constants.
the variance
zk's having
with near zero mean.
time histories
for thrusters
thruster
#7 was on from
run time of the test was
Table 2 contains
simdatexl
commands,
random
commands.
Table
Column
2 and 3 contain
2. Mean and Variance
9a-b
while
thruster
both outputing
show
the responses
from
Figures
lOa-b show
the command
#3 was on from
+/- 0.25 lbs force
10 to 13 seconds,
in amplitude.
values for both the CCS computed
1 identifies
the thruster
pair which
the mean
and variance
values,
of the open loop random
T.ht.ugrg..
#3 (CCS computed)
#7 (CCS computed)
#3 (simulation)
#7 (simulation)
of
since the
and computer
received
thruster commands.
Y-arJaag¢-.l 2i
-0.001538
0.001102
0.000998
0.005969
the random
respectively.
0.0062514
0.0069927
0.0066179
0.0076238
and
The total
15 seconds.
the mean and variance
and columns
of the distribution
a near zero mean distribution,
Figures
to these inputs,
#3 and #7, respectively;
5 to 7 seconds,
of a
c was set to zero, for both
corresponded
#7 and #8, respectively,
The products
of the Zk'S. For this
thrusters,
accelerometers
discrete
the mean of the distribution
to 0.5, while the variable
xk's also had a distribution
#3 and #7 were
output, in lbs, xk was the kth sample
a and b were both set equal
to the random
plots.
+ c)
of the Xk'S, dictated
of a and c dictated
#7 and #8,
equation
+/- 1.0), and a, b, c were scalar
of distribution
in a composite
#4, 2 lbs at 10 seconds;
can be seen on both accelerometer
was random
used
The total run time for this
test, the variables
which
shown
thruster
time histories
based on the following
thruster
and b, along
commands,
#7, 0.5 lbs at 20 seconds.
Zk = a * (b*xk
where zk was the kth sample
The same four thrusters
thruster #3, 1 lbs at 5 seconds;
type to be tested
to produce
The pulse
8a and 8b show response
The responses
The final excitation
to disturb the CEM.
were used here.
as follows:
thruster #6, -1 lbs at 15 seconds;
test was 30 seconds.
pulses
Although
the statistical
values
between
test and simulation
exactly.
This may be attributed to the differences
between
the EDS and the simulation.
During
the open
automatic
safety
loop command
shutdown
commands,
issued
commands
and stopping
features
by typing
the control
keyed
on the magnitude
assigne.d
a maximum
allowable
magnitude
exceeded
its respective
feature without
critical
exciting
per time sample,
phase, both the manual
shutdown
command
experiment,
commands
mentioned
above.
a zero command.
#2's critical
value was set to 0. lg; this experiment
for closed
Closed
Loop
Confident
transmitted
closed
8, but varying
damping
the vibrational
30 seconds
section
shutdown
damaging)
small
These
levels
degree,
were easily
artificially
reached
11 a shows the response
was stopped
using the sine
time history
of
The plot shows that the 0.1g
level
stopped and the thrusters
test, where
accelerometer
once this level was surpassed
indicating
that a sensor limit was exceeded
displayed
on the GSET.
system
and actuator
commands
were being
properly
and the CEM,
closed
of states,
were executed
of the CEM,
the CEM
looped
type involving
at
and the
was deemed
tests were started.
accelerometers
on the CCS/RIU.
and
Three
different
The performance
which
were described
for the first 7 seconds
of each run.
9
processed
#1-8 and thrusters
was the same for all three controllers.
long, and the same sine wave excitations
were used to disturb
were
In order to test this automatic
11 b shows the results
each a state feedback
motions
commands
CEM
signals
the CCS/RIU
in the number
signal
the CCS/RIU
the
loop controllers,
if any accelerometer
completed,
Tests
between
value;
of a similar
to begin.
that the sensor
was
Figure
loop testing
with
accelerometer
a critical value of 0. lg.
was satisfactorily
shutdown
Each
automatically
were correctly
safety
signals.
at which time the experiment
In both tests, messages
the open loop testing
The automatic
the test, all the thruster
Figure
were issued
ready
all the thruster
the critical
to the accelerometers.
at 4.5 seconds,
Once
in zeroing
was terminated.
was exceeded
terminated
were successful
critical value during
#1, which was assigned
experiment
shutdown
levels of the accelerometer
accelerometer
6.7 seconds.
Manual
in all instances.
signal level, called
and the
were tested.
the CEM to a high (and potentially
values were assigned
wave excitation
variables,
on the CDPP,
zeroed and the experiment
they did not match
in the random
of the CCS software
in "halt"
software
automatically
testing
were close,
goal,
of
Each test was
in the Open
After
#1-
a period
Loop
of 3
Tests
seconds
of fre.¢ decay, the controller
was then turned
run. The results for cach of tbe three
The first closed
order
loop controller
mass-spring-damper
virbrational
results
energy
(CAMAC)
show
comparisons
CCS/RIU
tests.
accelerometer
was executed
at 150 Hz and at 200 Hz, the
SSRL
VAX
For verification
3200 workstation
in the SSRL,
rack which
purposes,
performs
at 200 Hz; this VAX
the data acquisition
response
was
is used as the primary
Automated
and conversion
time histories
absorb
the same controller
and is tied into a Computer
of the accelerometer
Measurement
duties.
from both the SSRL
and
Figures
12a-p
VAX and
time histories,
from both tests matched,
data from the CCS/RIU
was noticeably
more noisy.
This was attributed to the fact
had 16-bit precision,
as opposed
to 12-bit for the A/D converter
A/D converters
have
16 times
per volt representation
versus
A/D's
in thruster
precision.
(i.e., 64 bits), while
command
the resolution
the RIU's
was limited
the CCS/RIU
the accelerometer
simulation
of the 16 state de.coupled
to 32-bit
time histories,
for the two tests matched
per volt.
the
13a-p,
to the SSRL
VAX was executed
precision
in
having
This in turn led to the
as seen in Figures
The control law on the SSRL
the CCS
although
of the RIU A/D, the CAMAC
204 counts
time histories,
when comparing
final check,
command
here.
actively
second-
from the CEM 4. This controller
factor to be considered,
histories
simulating
accelerometer's
slight differences
computational
digitally
which
thc RIU; the CAMAC
Another
paragraphs.
pair locations
Each
that the CAMAC
counts
controller,
of each test
at sensor/actuator
on the existing
Control
in the following
was a 16 state de.coupled
systems
real time control computer
3276
tests arc presented
of the latter test are presented
executed
on and left on for the remainder
for the two tests.
VAX,
is
in double
in its computations
from both tests, were fed through
precision
in the AP.
As a
a PRO-MATLAB
controller.
The resulting
simulated
thruster
their respective
CCS/RIU
and SSRL
VAX
command
computed
time
thruster
time histories.
During
the tests with the 16 state decoupled
CDPP
was tested
again,
manual
shutdown
command
controller
was working.
commands
and terminated
The second
was executed
controller
to ensure
controller,
that it can function
was tried on several
during
closed
In each case, the halt command
the control
to be tested
can be seen in Figures
controller
was considerably
on the CCS/RIU
accelerometer
14a-h and in Figures
better
shutdown
an actual
closed
loop tests, always
successfully
command
loop control
being
zeroed
from the
issued
test.
The
after the
out all of the thruster
execution.
at 200 Hz, and the resulting
histories
the manual
was a 42 state H** controller.
response
and thruster
15a-h,
respectively.
than the 16 state decoupled
controller
10
The controller
command
The performance
above.
time
of this
The final closed
loop controller
states.
This controller
thruster
command
respectively,
was executed
time history
were executed
operating
at higher
H.o controller,
It is obvious,
VAX.
from
these plots,
has also been encountered
on the SSRL
the present
at. The sampling
rate effect could
be noticed
when comparing
with the 200 Hz test results
for this controller.
Although
the performance
the 7 Hz instability
RIU
was more pronounced
Digital
Filtering
Tests
This section
discusses
the results
Analog
low pass filters
filtering.
The cutoff
in Table
3.
Table
the analog
filtering
is
previously
in
the controllers
is incapable
of
the 150 Hz test results
was poor in both tests,
tests which
in the RIU to provide
signal
analog
rate
freouencv
(Hz_
Cutoff
16.7
600
166.7
6000
1666.7
for the digital
filtering
prior to digital
RIU sampling
filter cutoff
60
filters were activated
were performed.
anti-aliasing
set by the selected
rate and corresponding
Samnlin_
Although
of the RIU digital
are automatically
3. RIU sampling
performance
in the 150 Hz test.
are also available
frequencies
VAX,
CCS/RIU
and
17a-h,
that the controller
this problem
rates, rates at which
response
16a-h and in Figures
This 7 Hz mode
To circumvent
sampling
this time with 60
at 150 Hz and 200 Hz. The accelerometer
has been excited.
tests with the SSRL
was another
plots can be seen in Figures
for the 200 Hz test.
poor, and a 7 Hz mode
control
which tested
rate, as shown
frequencies.
(Hz)
tests, they were not tested
by
themselves.
As mentioned
digitally
filtering
two pre-defined
in the System
Description
selected
signals
Finite
The first is the so called
sensor
Impulse
section
(FIR)
filter",
which
referred
as 'Triter #2", in this document.
means
no digital
filtering;
which
has a filter length
with a sharp roll off with little consideration
simply
to the CCS.
low pass filters
designed
which
the RIU contains
prior to transmission
Response
"structures
above,
given
to phase
a DSP board
Currently,
can be called
the RIU has
by the user.
of 110 (i.e., 109 states),
shift.
this "null filter" must be defined
was
This f'dter will be
A "filter #1" does exist and is termed
11
for
the "null filter",
since all sensor
signals,
whether
they are to be filtered
the CCS.
Figures
RIU sampling
18a-c
or not, are sent to the DSP prior
show
the frequency
and was designed
control
law computations;
hence,
the name
called
"filter #3".
19a-c
show
Figures
the 1553B
to
filter".
the frequency
smaller
pre-defined
phase
shift, for use with
For this document,
response
plots
filter has a
this filter will be
of filter #3 for the RIU
rates of 60, 600 and 6000 Hz, respectively.
To test the RIU digital
and connected
mimic
The second
to introduce
"control
down
plots of filter #2, for the three selectable
rates of 60, 600 and 6000 Hz, respectively.
filter length of 54 (53 states)
sampling
response
to transmission
filters, the RIU input channels
to a signal
the sensor
generator.
signals,
Fixed-frequency
sine waves
was sent to all eight RIU sensor
filters
1-3 described
channels
filters
were recorded
comparison
known
on the CCS's
of the RIU digital
input sine wave.
the results
of three
above)
channels,
were selected
GSET
of these tests are presented
involving
here.
Computations
different
for various channels.
for later analysis.
tests were made,
Control
however,
sensors
All the sensor
simulated
different
In these cases,
tests.
RIU digital
This post test analysis
filter outputs with PRO-MATLAB
Several
from the CEM
were then sent to the RIU to
as was done in the f'u'st part of the Dummy
The same sine wave
(using
were disconnected
involved
the
filter outputs of the
frequency
sine wave
inputs;
the RIU was set to sample
at
600 Hz.
Figures
generator
20a, 21a and 22a show
time history
sine waves,
compared
OUtpUtS havc
been
respectively,
scaled
The phase
effects
of the initialization
signals
shift became
Figures
outputs,
inputs,
more pronounced
= 0 in the plots).
20b, 21 b and 22b show
respectively.
"created"
There is a built-in
between
simulated
for the magnitude
signal
the filtered
to the differences
8.0 Hz signal
between
because
the RIU begins
delay
the time history
excellent
between
to process
to
RIU software
raw test sine waves
MATLAB
of the slower
however,
there are
data for the 8.0 Hz case.
This may
PRO-MATLAB
used in the actual
simulation.
could not be used directly
sampling
12
plots of RIU filter #2
agreement,
the actual raw 8.0 Hz signal,
used in the corresponding
because
The group delay
not when the CCS begins
five second
test data and simulated
filter tests, the GSET-recorded
simulations
frequencies.
of the digital
outputs of 0.1 Hz, 1.0 Hz and 8.0 Hz sine wave
The 0.1 Hz and 1.0 I-Iz tests show
between
Note that the RIU
on the GSET.
comparisons
and filter #2 PRO-MATLAB
be attributed
at the higher
of the digital filter was not recorded
and the start of data recording
larger differences
to account
in its DSP as soon as the RIU is fully configured,
record the data (dme
configuration
to RIU filter #2's outputs.
by the dc filter gain of 0.749,
filter.
the sensor
plots of 0.1 Hz, 1.0 Hz and 8.0 Hz raw signal
rates of the CCS.
test, and the
For all of these
in the PRO-
The RIU was sampling
thetest sine waves
However,
at a rate of 600 Hz since the digital
the CCS cannot
the test data at 150 Hz.
test sine wave
sine waves
to match
times
sampled
were based
upon
signal
generator
outputs
#3.
points
did not match
led to the slight differences
Figures
the measured
filter outputs
in PRO-MATLAB.
the created
amplitude
at 600 I-h:,
These
created
sine waves
and phase,
were made
but contained
In the 8.0 Hz case,
4
period
of time.
signal
as well as in the 0.1 Hz and 1.0 Hz cases,
in the filter test data and the PRO-MATLAB
the created
filtered
sine waves,
to RIU filter #3's outputs.
respectively,
scaled
compared
by the de filter gain of 0.913,
comparisons
between
simulated
One interesting
the time history
outputs
of 1.023 higher
true for all three frequency
Standalone
Tests
As previously
mentioned,
test cases,
CCS out of the loop, although
commands.
plots of RILl filter #3 outputs,
than other
presented
inputs,
respectively.
tests was that the output
of RIU
with the same outputs.
This held
control laws on its DSP with the
it is possible
for the EDS to initially
"standalone"
mode
PC serial interface.
These
control
sampling
rates
pairs, can be specified
and filter #3 PRO-
to execute
via a standard
Only
23b, 24b and 25b
the RIU has the capability
are entered
he processed.
filter
there
i.e., in the form of digital
RIU digital
both in order,
of digital
At present,
in the DSP; the coefficients
laws are limited
filtering
RIU channels
programming
of the three available
Figures
that the R1U
above.
This is the so called
form of transfer functions,
less for this filter.
during the digital
again,
for the magnitude
of 0.1 Hz, 1.0 Hz and 8.0 Hz sine wave
point discovered
#1 was a factor
shift is much
to account
Note,
600
data.
plots of 0.1 Hz, 1.0 Hz and 8.0 Hz raw signal
MATLAB
RIU
the RIU digital
andre, cording
23a, 24a and 25a show time history
have been
channel
of frequency,
over a given
Also, it can be seen that the phase
show
simulate
raw test sine waves;
in terms
for that rate.
C) and was sampling
at 600 Hz, had to be created
the actual test sine waves
Hz sampled
to properly
the measured
of sample
were designed
at 600 Hz (see Appendix
Thus, in order
signals,
the number
which
operate
filters
three different
control
with the present
RIU.
13
with open loop
of the RILl. The RILl control
laws take the
Infinite
Impulse
in on the GSET
Response
of sensor
functions
functions,
inputs
(IIR) filters,
and transmitted
- 60, 600 or 6000 Hz.
transfer
of this.
excit the CEM
law transfer
and in the number
is no explanation
and actuator
i.e., three
to the RIU
must be designed
Currently,
different
for
for one
the RIU control
outputs
which
actuator/sensor
can
With the above
standalone
limitations
mode.
commands
were transformed
using available
an appropriate
A six state decoupled
and 8) and three thruster
matrices
in mind,
routines
controller,
(thrusters
into 3 separate
accel.#1
in successful
The above
-2.3730"10
GSET,
loop tests on the SSRL
transfer
functions
particular
excited
control
Subsequent
off.
Once
law was encountered
tests, with the RIU actuator
that the RIU computed
problem
has been discovered,
coefficients,
which presented
Other
RIU standalone
tests, involving
successfully
programmed
into the DSP onboard
zero.
tested.
These
tests used the pre-defined
as transfer
functions
for 3 sensor/actuator
the CEM
with this
the thrusters
never
fired.
on an oscilloscope,
Although
no direct cause
of the
of the
for the RIU.
functions
digital
pairs.
mode,
to the small magnitudes
difficulties
other transfer
14
testing,
used was the
A problem
signals being monitored
has been attributed
numerical
of the CCS
in standalone
the CEM.
in the RIU, in repeated
command
verified
the RIU with the GSET
tests, the only component
actuator values were always
numerator
-2
at 200 Hz, wcrc previously
the RIU was executing
the problem
z-2
3200.
to see if the control law could control
revealed
been
VAX
for the RIU standalone
-l +0.9930z
for execution
were programmed
all the rest were turned
was manually
-3 z-I +2.3623,10-3
1.0000-1.9930z
closed
via a RS 232 interface;
1.9940z -1 + 0.9943z -2
1.0000-1.9924z-I +0.9927z-2
law, digitized
six state
* 10 -4 z -2
accel. # 3
of this control
The orginal
# 1, 3
at 600 Hz and are shown below:
-3.0964"10-4 z-I +2.9472"10-4 z-2
accel.#8
functions
inputs (accclerometers
thruster#
3
thruster#8
Transfer
* 10 -4 z -I + 2.3344
1.0000-
to test the RIU
order, single input, single out transfer functions,
they were discretized
2.4257
thruster#I
with three sensor
# 1, 3 and 8) was selected.
second
in PRO-MATLAB;
control law was choosen
programmed
FIR filters
in the DSP, have
available
in the RIU,
Development
problems
This section
outlines
and
recommendations
the problems
and some of the solutions
which
encountered
during the development
had to be implemented
effectively
as possible.
either
being
in the CCS (i.e, in one of its components),
CCS
Problems:
-
_.Q.,L]dSI].Q_ -
Both software
How and where
and hardware
The initial plan called
for using
the data on the 9 track magnetic
transfer
the test data to a PC for post test analysis.
complicated
software
tape.
which,
1553B bus and storing
the data onto its hard disk.
disk space
which
method
was further
enhanced
would
"talk"
was ample
record
all data transmissions
to the 1553B
or in the RIU;
they were as follows:
However,
in favor of simply
HRM
I/O interface
this would
to capture
offthe
1553B
from Remote
mode,
bus.
where
have
entailed
writing
time of the CCS.
the test data directly
off of the
25 megabytes
of
test runs.
This efficient
Terminal
mode,
the GSET
This doubled
of the GSE and
then be used to
the delivery
to hold data sets from multiple
bus to Bus Monitor
would
The hard disk had over
the GSET
as
data.
have delayed
using the GSET
by switching
as
here are categorized
The IEEE 488 interface
in turn, would
function
system,
reported
the existing
This plan was dropped
unused
problems
the CCS/RIU
to store the real time experimental
storing
additional
to make
of the CCS/RIU
would
in which
just "listen"
it
and
the speed of real time data
recording.
- The
1553B interface
card in the GSET
transmissions
off of the 1553 bus, which
single
values),
sensor
or entire
resulted
data transactions
PC had a tendency
in either
single
(all actuator
properly.
Fortunately,
only 1-3 data recording
occurred
over a test run containing
syncs.
However,
increase
the number
at the higher
_,_).ld2.1_[.-
errors
sampling
The problem's
GSET
was of an old design
extra
funds
improved
of data recording
card with better
associated
recorded
or
with a
occurred
data for some
5000
in a test run tended
to
to the fact that the 1553B interface
performance
a better
limitations
were known.
1553B card (SCI planned
speed performance
15
values)
actuator
data drop outs were rare; on average,
which
was attributed
to procure
(single
the data
time
200 Hz.
from SCI, and its speed
nor time were available
1553B interface
errors
rates, i.e., above
cause
these
data words
or sensor
time sync not beingrecorded
of misreading
later in 1991).
to market
card in the
Neither
an
A post processing
software
fix was introduced
was completed,
the file containing
Each data set associated
any large deviations
detected
to help alleviate
among
this problem,
the recorded
data transactions
with a time sync was compared
in data indicated
the data of a given
and worked
missing
words
as follows.
was checked
After a test run
for data errors.
with data sets of previous
or whole
time sync, the software
transactions.
time syncs,
If any error was
simply
copied the appropriate
At the end of this post
transaction
(be it sensor
or actuator)
from the previous
time sync.
processing,
a file was made available
to the user which
listed the errors, thei