via a fiber optic link, where data was then transmitted through the Tracking and Data Relay Satellite. Sys- tem (TDRSS). The received data acquired at the. White.
N95- 17186
i_:i _,'(::'ikL
::._
LI
.:
: W¸= • ...:.
:ii_!il _i!_i_iiL _
•
+
.. (")
il ) ...... e;
END-TO-END LENGTH
COMMUNICATION
PACKET
TEST
STRUCTURES
ON VARIABLE
UTILIZING
AOS
TESTBED
/iii+i:L
W. H. Miller*,
V. Sank +, W. Fong*,
Folk**,
B. Conaway
+, K. Michael*,
* Goddard i_?::
J. Miko*,
Space
M. Powers*,
J.
P.-S. Yeh*
Flight Center
.
+ Fredrick ** Advanced
II!:_IIL •
Technology
and Research
recommended
This paper
tecture (AOS)
lessly tem.
describes
a communication
demonstrated
the
compressed images These compressed
formatted mittee
into variable for Space
Data
test, which
transfer
of loss-
in an end-to-end images were
length
Consultative
Systems
(CCSDS)
sysfirst
ferred
from
the
Simulations fiber
link,
tem (TDRSS). back and This
Sands to the
Compackets
Complex
decompressed
back
describes
data
acquired
the data
a
Sysat the
the advent space
data more
instruments
data rates national in order
complicated
and formats.
components in order
achievable
speed
which
components, link charac-
variable
to provide
the international
CCSDS
57
between
commercial
plat-
data
for the
AOST
AOS.
The
is implemented
in
insight
limitations diagram
regarding
for actual in Figure
including
net-
of this (GSFC)
flight
1 shows
an instrument
sim-
channel
transfer
is capable
frame
of implementing
the packet
specified
in the .standards
in Figure
2. A salient
book
feature
of
is the ability to transport varipacket, as opposed to the con-
fixed-length allows
generator.
packet
packet data
to be multiplexed
structures.
from
in a much
This
different
instru-
more
flexible
way in the data system. For :data originating from one single instrument, the variable-length packet is also length
a natural bit string
structure resulting
for holding from
ing fixed-length instrument scan line of image data. a
archi-
ulator followed by a packet generator, a high-speed multiplexer, additional instrument simulators, and a
ments
inter-
to issue
block
the data architecture able-length CCSDS
are
to handle
data
to provide and
and re-illustrated
satel-
generate
The desire
The
architecture
in order
data system
a testbed
these key components
testbed
systems
and
of the
hardware
data
scientific
space
stations
developing
The
cross-support across different platforms to better utilize the science data globally
has prompted
been
key
structure
communication
advanced
has
ventional
of sophisticated
becoming
ground
images.
H L_
With
on space
to transport
algorithm,
INTRODUCTION
lites,
forms,
virtual
captured
original
the compression
flexibility
hardware.
transferred
was
the AOST configuration, key flight data formats, and the communication teristics and test results. 1.0
Satellite
was
to the
via
then transmitted
Relay
(WSC)
where
Frequency
(RFSOC),
was
received
AOST
paper
data and Data
The
Radio
Center
where
the Tracking
White
to the
Operations
optic
through
AOST
standard
specified in the Advanced Orbiting System Blue Book [1]. This architecture provides
works. To demonstrate the capability architecture, Goddard Space Flight Center
in the Advanced Orbiting System Testbed (AOST). The CCSDS :data Structures were trans-
: ili__ •
and Associates
Abstract
successfully
!,? ii_ ,
Herold
the variable-
losslessly data,
such
compressas from
a
In parallel with AOST effort, GSFC is also engagedin the developmentof datacompression technology.Data compressionprovides a viable meansto alleviate the demandson onboardstorage,communicationsbandwidth, station contact time andground archive requirements.There are two types of data compression: a losslesstechnique,which guaranteesfull reconstructionof the data;anda lossytechnique,which generallygives higherdata compactionratio but incurs distortion in the reconstructeddata. Losslesscompression generally results in variable length compressed datadue to statistical nature of the original data. To satisfy the many sciencedisciplines, lossless data compression has become the priority for development.After extensive research,the Rice algorithm [2,3] was chosen and developedinto hardware.In 1991,a hardwareengineeringmodel wasbuilt in an application specificintegratedcircuit (ASIC) for proof of concept.This particular chip setwasnamedthe UniversalSourceEncoder/ UniversalSourceDecoder(USE/USD)(Venbrux, 92)[4]. Later, it wasredesignedwith severaladditional capabilitiesandimplementedin Very Large ScaleIntegration(VLSI) circuits using gatearrays suitable for spacemissions.The flight circuit is referredto asUniversalSourceEncoderfor Space (USES).The fabricated USES chip is capableof processingdata up to 20 Msamples/secondand will take dataof quantizationfrom 4-bit to 15-bit [5]. In the following sections, we will provide a brief
description
rithm, AOST
of the
the overall and physical
2.0
THE
data
compression
communication link characteristics.
LOSSLESS
fixed
imagery
Fig. 3. It consists
of a preprocessor
data samples and subsequently symbols suitable for the entropy This
entity
parallel yielding selected.
is a collection
over
a large
the least This
number
selection
of coding
is performed
the
tested
3.0
SYSTEM
3.1
End to End System
The
on various
system
is linked
which
is depicted
via an optical
transmits
the packetized
and
to the
later
transmitted
communication
4.
to the RF data
to the
a TDRS on a at the WSC
AOST
via
NASA
(NASCOM).
Source
3.2.1 Data
in Figure fiber
(WSC) via was recorded
The
instru-
Description
White Sands Complex Ku band carrier. Data
3.2
on typical
DESCRIPTION
end-to-end
The AOST SOC,
codes
of as
[7].
Equipment
Source
source
data
can be either
simulated
instru-
ment data or a video frame of data acquired from a CCD camera. In both cases, the data is first loaded into a frame buffer before each scan line is passed
to the compression
porates
the USE
is then passed cessing.
chip.
hardware
Each
which
compressed
to the packetizer
Packetizer
packetizer
encapsulates
and
is shown
in
operating The
incor-
scan line
for further
pro-
Multiplexer
System
data
over
a fiber optic at a burst
from
the
packets
transmittertransfer
instrument, [8], and sends
receiver
inter-
rate of 80 Mbps.
A
separate packet is formed for each video scan line with the segmentation flag in the packet header used to treat an entire video frame as a large block. The segment flag is set to "beginning
in
segment"
option
for the first scan line of a video
is set to "continuation
bits will be over
takes
it into CCSDS
face (FOXI)
map them into coding module.
range.
input
to signal
Description
to decorrelate
of options
entropy
by the
is used
of Huffman
[6] and has been
ment data
The
algorithm
determined levels,
that of a collection
3.2.2
the
ALGORITHM
of the Rice
of bits,
quantization
selected option for the block. The performance this algorithm has been shown to be the same
them The architecture
number
sample
adaptability field of a
algo-
system,
DATA
COMPRESSION
of J, typically 16, samples to achieve to scene statistics. An identification
a block
58
segment"
scan lines;
and it is set to "end
final video
scan line of a video
data of
frame;
it
for intermediate of segment" frame.
Frame
for the syn-
_;_,_i _!! ¸!il i! i _!_!II_!'
i_!!i_ii_
chronization is derived the FOXI interface. The
multiplexer
service CCSDS i:_i _
?ii
!i!ili_ ii_ i_,i ?i!i¸
:i •
WTFF.
where through
Access
without (VCA)
and
in
1) path
the multiplexer to the Wideband
to the output
on availability
processing and service mode
channel
a round
proto the
is granted
robin
ing
function.
In general,
rate for a channel, and grants
access
hardware 3.2.3
the more
are given
to be data
rate
are provided
driven.
of
Details
of the
in [9].
The
ence
ii
system
is designed
to serve
1, for up to seven
transition,
the pseudo-noise
is utilized.
user virtual
before
being
as necessary
and
are output
Grade-2
channels
service (RS)
to maintain
on a single
3.3
Data
channel
error
from VCDUs
Virtual and
Channel:
formats
or MPDUs,
it into
virtual
channel
at rates
composed
of five interleaved
bytes
255 bytes (10,200
ing check
up to 100 Mbps. each.
bits)
symbols.
long,
circuits to (CVCDUs
with a frame access
synchronization data
unit (CADU)
was
data
of the pattern to the I
synchronization and is neither
received
transferred
patRS
at the WSC
to a workstation
on the I and pro-
with software tools. The Q sent to a communications bit
test set for real
a hardware
time monitoring.
virtual
RS decoder;
a software
channel
and packet
data
decompressor;
display
package
words
RS
detection
soft-
and a soft-
for the workstation.
original
Detector data
as organized
bits in each byte
byte
boundary
is lost.
searches
the received
basis to find the frame ensure that it is the
are
sequence
con-
(32 bits),
59
and knowledge
The
frame
of
detection
file on a bit by bit
synchronization frame marker
bits that happened
marker,
by the
the file position
marker. To and not a
to be the same pointer
was
moved one frame length from the initial marker and data was examined for a second marker. In all a
which
of data
as the frame
into frames
forming a data frame from the WTFF, the
are serialized
(i.e.,
are appended
is created,
Frame
software
the RS encod-
pattern
PN
Equipment
data
is thus 1275
CVCDUs
byte
a stored
the multiplexer
WTFF is a series of bytes structure. Once transmitted the
user
frames
CVCDU
including
When
The
correcting
? • •
channel
Capture
a software
3.3.1
line.
The frames
RS code
Each
gener-
a preset
Fig. 2).
A VC unit receives
CVCDUs) taining
sent through
rate (BER)
ware;
by including
error
it is, each
with
by the PN generator.
encoder;
from
serial
is provided (255,223)
transition
bit
The capture and analysis equipment is composed of a 32 Mbyte solid state memory connected to a Sun workstation via an ethernet; frame detection
(VC)
with frames
code in each of the eight virtual channel form coded virtual channel data units formed
When
or changed
ware image
rate
(PN)
coded
are padded
data
To ensure
The frame separately
frames.
frames
set by hardware.
or Q data outputs. tern is generated
software;
These
controller
re-configura-
Generator:
is XORed
Data messages arriving from VCs are buffered and then standard format data transfer
a Reed-Solomon
_i! !
Transition
plus an idle channel. any one of the user inserted into CCSDS
CCSDS
_iili ¸¸• •_
as a
providing transfer frame generation using of the AOS services, as defined in Refer-
the idle channel !_ii _ _iiiiili_
[10]
system
system
ID (VCID)
PN Code
cessed predominantly channel signal was
WTFF
by the
or during
Data can be received as a fixed length data unit (MPDU in VCA service) or as CCSDS packets (Path Service).
All packetized
(WTFF)
gateway a subset
initialization
data streams.
caus-
Transfer Frame Formatter
Wideband
VC is configured
upon
CVCDU
the
the number
to that channel,
over I or Q output
Each
ator
polling
the higher
can be transmitted
tion and has a unique
passes Transfer
sequence. This polling occurs once every 400 ns, which is rapid enough that it results in a statistical
requests
7i I
signals
where the multiplexer produces multiplexing tocol data units (MPDU) and transmits them
packet
_
control
in two modes:
Frame Formatter (WTFF) 2) virtual channel access
multiplexing
ii_i_ _ _
operates
mode packets
based _i!!_•
through
of the data collected,
a bit slip (or bit addition)
was
neverobserved,makingthe needfor amoreelaborateframe synchronizationstrategyunnecessary. 3.3.2
Reed
After
byte
decoding
alignment
and
frame
Each
detection,
frame
RS
output
from
the RS decoder has a 16-byte status block appended to it. During the data flow with the compressed
variable
length
image
packets,
the error
rates were never severe enough to cause uncorrectable frames. During the portion of the test where the purpose was to evaluate the Ku band physical link using the CCSDS structure, frames that were found
to be uncorrectable
were examined the correctable frames data ated
the data portion
by computer
using
structure, and the parity by re-encoding the data.
were
also retained
raw data determine 3.3.3
not deleted.
of the frame knowledge
portion CCSDS
and RS decoded.
and
Packet
was
of the
was generidle frames The received
was compared with RS corrected burst error statistics.
Channel
and
data
to
pressed
channel
output
identification
found.
Compressed
(VCID) video
particular
application
particular
virtual
file
for each number
packets
were
identification
channel.
The
PACKET
was run on the VCID file of interest separate output files for each APID
was
assigned
(APID)
block
to convert
algoat the
the data back
format. The resulting image file was then accessed
display
software
TEST PARAMETERS ANALYSIS
4.1
Test
Testing
chip,
package.
AND
LINK
variable
length
Parameters
of the AOST
using
packet
structures was performed by using losslessly pressed images. The first image transmitted
comwas a
prestored Landsat image. After the Landsat image was received, real time images from the video camera and packetized data from the simulators routed
through
on the
the link ured
BER.
equipment
and
I channel at 80 Mbps was not used. PN data
the
source
while trans-
RF Q channel The
WTFF
to run at a continuous
was
used
I channel output
to monitor was
config-
rate of 80 Mbps
a Virtual nel #
on a
program
variable
Chan-
Virtual Channel Mode
Data Source
Data Rate
"1'
Path Service
Simulator
16.0 Mbps
2
VCA Service
CCD
-28 Mbps
3 63
Path Service Idle channel
Simulator WTFF
Link
camera
16.0 Mbps as needed
ANALYSIS
Decompressor
The packet data
(USD)
512 sample line as well at the start of each com-
4.0
4.2 3.3.4
16 sample
by a commercial
and produced found in that
VC. The APID file of interest contained length packets of compressed video data.
performed
(including idle channel fill frames) by using an 80 MHz crystal on its high speed output board. The WTFF configuration was as follows:
virtual that
decompression
Decoder
compressed information
output on the WTFF the WTFF Q channel
Detection
the
the operation
Source
to its uncompressed 512x512x8 bit binary
mitted
a separate
Universal
start of each as the header
were
The file created after RS decoding was then processed by a software program called CHANNEL which examined the CCSDS frame header and produced
performed
simulated
realizing the lossless data decompression rithm. The routine used the reference sample
They
for burst error statistics along with frames. When uncorrectable
occurred,
corrected
were
routine
algorithm by the
Solomon Decoding
was performed.
decoding
was
containing
file associated
further
broken
one compressed
512 variable-length one fixed-length
lines image
with down
the image
flame
into a separate image
frame.
file
These
were decoded to generate frame. In software, this
In addition to evaluating compressed variable length packets, this test allowed an examination of il;
the channel characteristics of the K-band Single Access (KSA) return link through TDRSS. The primary objective was to evaluate a BCH code
i!
proposed
::4:
for a LANDSAT-7
300 Mbps
return
link.
fi!
i::i
60 .::::
It was important
••:!•i ¸i•¸_• ii!i!iiiii,!_ili/;'_i
to understand because
the error
character-
istics
of the channel
the proposed
binary
BCH errors
code (1023,993,3) can only correct in a block of 1023 bits. This section
3 bit will
identical to the digital was found to be true. 5.1
Image
concentrate on the performance of the link used in this experiment in terms of BER vs. Eb/No. Link
Since
analysis
independently
was performed
pattern
of NRZ-M
by transmitting
data
at 300 Mbps,
a PN data QPSK
modu-
lated (150 Mbps on I, 150 Mbps on Q) through the KSA return channel. The data was recorded in one minute samples, and six pairs of C/No and BER measurements were taken at WSC with 223-1 and 27-1 PN coded data. The measurements are summarized
in Table
1. Using
noise density, the required ated power (EIRP) values TABLE
1.
C/No,
Measured C/No (dB)
i_• iiii!ii_iii! _• •
the measured
carder
to
effective isotropic radiwere also calculated.
BER and RF SOC EIRP Measured Bit Error Rate
Data Points
EIRP Required for Measured C/No
No streaks
error Loss
the
of the original.
or drop
outs
occurred
compression
in the images.
technique
to each
scan
line,
was
several image observed.
lines,
5.2
Performance
Channel
but
no
One
of the concerns
was
errors
on the TDRS
Ku band
applied
a decompression
would be expected to corrupt of a user data CCSDS frame
examine
This
Quality
an entire line. would impact
corruptions
were
to evaluate
all the parameters
link.
the burst
A test cannot
that vary
over
a satel-
lite's lifetime, but it can at least provide a snapshot of some of those parameters. In order to simulate an end-to-end
system,
95.1
0.8e-3
51.2
95.3
3.5e-4
51.4
96.5
9.0e-5
52.6
WSC and retransmitted fmally decoded.
98.9
2.0e-5
55.0
During
99.3
3.0e-6
55.4
ing (NRZ-M)
99.9
2.0e-7
56.0
the data
the link portion was
The disadvantage
was recorded
to GSFC
at the
before
being
of the test, differential
used
to avoid
of NRZ-M
data
cod-
inversion.
is that it causes
each
A plot of the six data points are shown in Figure 5 along with the ideal BER vs. Eb/No curve. The
error to appear as two errors which may or may not be consecutive. In the data observed, no errors
separation
greater
from
measurement, of the
the ideal
about
data
points.
implementation the required eters were
curve
an average This
value
varied
with
each
consecutive)
was
Landsat
loss. It is important
taken
as the
to note that for
EIRP calculation, several loss paramassumed based on estimates and the was cloudy
settings, lengths
RESULTS
Since
the WTFF
used
the CCSDS
recommended
(255,223) RS code, it was expected that as long as the BER was about 1"10 -4 or better, the RS decoding would expected error
free
correct
all of the errors.
that the decompression and
the
reproduced
It was therefore process image
would would
be be
?i __i::
61
2 bits in length were
7 power where were
(errors
observed level
when
was
the BER seen
used.
was
(Table
were the
greater,
2). For
power
longer
than 11 consecutive correct starts and ends with an error.
bits.
A burst
burst
of incorno more always
BER vs. Error Burst Characteristics
BER
Max Length Burst
2.2e-5
2
2
1.3e-4
14
4
8.5e-4
22
6
The CCSDS
proposed
this analysis,
as a group there were
2.
always
At lower
burst was arbitrarily defined rect and correct bits where
TABLE
5.0
than
of 3.6 dB fo_ five
weather conditions of that day (which with light rain) at the transmit site.
:?il _
version
frame
sequence
of
count
Max Errors Per Burst
was continuous
with no gaps. Bit slips (or additions) were never observed in this or. other AOST tests with the TDRSS.
a
?!ii(:::ii! i:i:i_
H
i:,
:
6.0 ii
CONCLUSION
The
ii_} _
i !ililI _
CCSDS
recommendations
architecture
have
been
compressed
data
being
separate
for
multiplexed channels.
images
AOS
put to a physical
instrument
compressed :_:i_iii!_ {!,
4. Venbrux,
were
data
test with
Losslessly
5. MRC,
decom-
cause
long
of the decompression
streaks
in the recovered
images
operation.
by the variable
length
ing from lossless compression. TDRSS link contained errors purely
thermal
transmitted
(random)
6. Yeh,
from
the
as
packets
This
result-
for data analysis
is
based on statistics gathered during a short period, therefore no statement can be made about the • : ::i ¸
burst
i__/i_ i
i:?
I
environment
that would
the TDRSS antenna of the Earth.
toward
technical 8. Packet January
when
other areas
Flight
nical tion,"
Data
Links:
sultative CCSDS
ii!
?
Architectural
Committee 701.0-B-1
Available
from
nications
and
TS),
National
istration, 2. Rice,
Systems:
CCSDS Data
systems
Commu-
Division
(Code-
and Space
Admin-
DC 20546,
"Some
Con-
Data Systems, October 1989.
Secretariat,
Aeronautics
E,
and
Specification,
for Space Blue Book,
Washington,
Robert
Networks
practical
USA. universal
noiseless coding techniques," JPL Publication 79-22, 1979. Available from author, JPL, 4800 Oak Grove Drive, Pasadena, CA. 91109, USA. 3. Rice,
Robert
Miller,
E, Pen-Shu
"Algorithms
Noiseless Computing Diego, CA,
Yeh and Warner
for High-Speed
Coding,"Proceedings
H.
Universal of the AIAA
in Aerospace 9 Conference, Oct. 19-21, 1993.
the
1993.
E Rice
Optimality
Coder,"
paper Telemetry
and
Warner
H.
of a Universal
Proceedings
of the
AIAA San
San
62
No. 3441,
1993.
Blue Book
CCSDS
102.0-B-2
1987. Wai, "AOST Technical hardware description," Center,
10. Betancourt,
Orbiting
-
7. Yeh, Pen-Shu and Miller, Warner, "Application Guide for Universal Source Encoding," NASA
References 1. Advanced
for Space
Computing in Aerospace 9 Conference, Diego, CA, Oct. 19-21, 1993.
9. Fong, SEMP
be observed
is pointed
Vol. 2, No.
Encoder
Mexico,
Robert
"On
Noiseless
The Ku band consistent with a
GSFC.
Technology,
Source
of New
Pen-Shu,
Miller,
Therefore
environment
N. Liu,
Lossless on Circuits
Preliminary Product Specification, 2.0, Microelectronics Research Center,
University
one can say that the data generation and recovery system worked as expected. No problems were introduced
"Universal
USES," Version
of compressed data is very senerrors which, if they occur,
results
Yeh and Muye
for High-Speed IEEE Trans.
data
and
pressed without any distortion. The achieved compression ratio is about 1.8 for the Landsat image. This type sitive to channel
Pen-Shu
and Systems for Video 4, Dec. 1992.
with several
received
Jack,
"A VLSI Chip Set Data Compression,"
Description: Goddard Space
1994. Jose and Folk,
John,
"AOST
Tech-
Description: WTFF hardware descripGoddard Space Flight Center, 1994.
Video
VC Input
CCD
Camera 1 Buffer Frame
Compressor
[--7 Multiplexe
Computer File
_Data
LJ
Packetizer_
I Is_SJu_ent_
Packetize
0_'_ 1
Frame Formatter (WTFF)
]
RF SOC
Packet Simulator
i_
Packet
|
_i_iiiil _ Simulator_,---_
Figure
packetID I segcnI packetlen I
data field (variable)
.._
16
16
packet
16
CCSDS
1st header
p_
I (k+l)th
kth packet
packet
MPDU packet
(k+2)th
]
MPDU
zone
J
I frameheaderI VCDUdata unit zone
_r]
CVCDU 32
8872
sync ] marker I
CVCDU
I sync I I markerl
10200
Figure
2. CCSDS
63
Data Unit Structure
CVCDU
CADU
:,i iJ
iiii? ii:::il i_('_: ¸
/ii i_! _'• i!iiii!ii: •
Figure _i_ ii_!
3. Rice Compression
Algorithm
i
TDRSS
°u ceI Capture
"_--'---NASCOM
J
¢
f
Figure
4. End to End Test Block
100 .
i
,
•
,
i
,
,
0 _
10-2
Diagram
,
i
,
,
,
i
,
•
,
meosured /dee/ PSK, no cod[n,
w 10-4
m
0 10-6
o 0
10-8 i
8
i
i
10
i
,
i
12
Eb/No Figure
5. End-to-End
64
Test BER
vs. Eb/No
i
i
I
14
i
i
r
16