Diary
# 24
EXPLORING
SATELLITE OPTIONS
There
are many places on this earth where no 'terrestrial' wireless
(or wired) links to base stations, universities, or the Internet
are feasible - or even possible. Some are even impracticable
to reach - or service - manually. Yet many such areas need to
be the subject of biological scientific or long term environmental
research projects. And in the future, with very large scale,
real time, comparative analysis of data from nature being made
possible by high speed internetworks linking far flung data
bases, the need for deploying more data-collection points will
go up dramatically. Collecting
the information from them will require more and more automation
of the task.
One
theoretical solution is to use a bi-directional satellite Internet
service to reach an independently-powered base station, and
then reach out by terrestrial wireless to various data loggers
within the radius of range of these data radios.
But
satellite Internet services, smack up against the iron limits
of the speed of light, introduce latencies that can range from
500 to 2,000 milliseconds when the signal travels up to 44,000
miles up and down, as well as unpredictable delays through network
congestion. All of which can affect the interactive timing of
data loggers which are not specifically designed to endure variable
rates between software. Including one of the most popular and
reliable products - Campbell Scientific's PC208W packages and
their CR10X, 23X data loggers in the field far away.
Wireless
connections, between the data logger and the base of the satellite
services, introduce still other variables which have never been
examined closely, which we will attempt to do over the next
two years of this project.
Of course
the object of all this is to use a combination of wireless solutions
to get the most reliable, flexible, and useful network at the
lowest total cost.
First
the Satellite Link
The first
thing was to install and test a late model Satellite Internet
Service under controlled conditions. We chose Tachyon's recent
deployment of a 64-256kbps (up) to 300kbps-2mbps (down) service
which uses SatMex5 KU band satellites to start with.
Red5,
a San Diego based Internet service, which partners with Tachyon
provided the installation on our Old Colorado City Communications
at 38 degrees 50 minutes 51 seconds North Latitude, 104 degrees
51 minutes, 40 seconds W Longitude, 1873 meters elevation. A
one meter dish antenna was installed on the 3d floor roof of
our 1891 Victorian Brick building, with the de-icing option
installed. Three wires were extended 100 feet down an old chimney
as a conduit to our Lab premises. The IDU was installed inside
our office space. The system was turned on July 17th, 2000.

Tachyon
1.1 Meter Dish Antenna on Old Colorado City Communication's
Rooftop
ODU
- Outdoor unit - Tachyon Radio

IDU - Indoor Unit of Tachyon Satellite
in Old Colo City Communication's Lab
Old
Colorado City Communication Lab
The normal
installation by Red5 gives the customer 5 IP addresses to use
off the 10baseT RJ45 port on the IDU.
The first
thing to check was its performance from a Windows machine attached
to the IDU. Running standard Explorer Web Browser was quite
satisfactory. Which was to be expected, for the whole service
seemed to be optimized for the office/browser market. Pinging
ICMP packets to various remote destinations - including looping
around from the Tachyon base facility in San Diego, through
Red5 to the Net, thence back to our Old Colorado City network
via Qwest and our Aironet wireless backbone from Qwest to our
offices, and into our Linux servers, resulted in measurements
from 800 to 1200ms latency.
I observed
an odd phenomenon. When pinging the IDU from our Linux systems
we noticed an alternating patterns of delays. Every other returned
packet would be up to twice the delay of the other. i.e. when
pinging the IDU address from our Linux server, over the national
Internet results in this typical pattern.
PING
216.7.145.65 (216.7.145.65): 56 data bytes
64 bytes from 216.7.145.65: icmp_seq=1 ttl=235
time=1620.3 ms
64 bytes from 216.7.145.65: icmp_seq=2 ttl=235
time=621.3 ms
64 bytes from 216.7.145.65: icmp_seq=3 ttl=235
time=1305.9 ms
64 bytes from 216.7.145.65: icmp_seq=4 ttl=235
time=892.4 ms
64 bytes from 216.7.145.65: icmp_seq=5 ttl=235
time=1304.9 ms
64 bytes from 216.7.145.65: icmp_seq=6 ttl=235
time=882.7 ms
The pattern
was the same from other systems where we have an account, such
as one in California that is directly connected by DS3 to the
backbone of the network.
We are
not sure what causes this, and wondered whether it would cause
erratic behavior when the PC208W software was operating. It
appears to be specifically a characteristic of the Tachyon satellite
portion of the link.
Satellite
to Server Performance
The next
step was to install a small Red Hat Linux powered IBM Laptop
as a web server connected directly to the Tachyon IDU.
This
configuration might be used in the field, if remote sensors,
or data loggers, or even video or audio devices are deployed
which send 'back' only as far as the web server their data.
While it is becoming possible and sounds very appealing to permit
anyone on the Internet to reach, in real time, data loggers
and other devices deployed in the field, that may not be a good
idea in most cases. Small devices can be overloaded by simultaneous
accesses to the small ram onboard. Preliminary screening of
data, to reject the obviously bogus data from failed sensors
or broken connection, will often be desirable. So permitting
near, but not actual, real time access to the data, images or
sounds, can be done by programming systems to automatically
send, store, and updated by automatic feeds from the end devices
- whether they be numerical, video, or audio, becomes possible
with a small, low power drawing Linux web server installed at
the end of the satellite feed, between it and the data logger
or field sensors, such as the Axis cameras we have begun to
experiment with.
The
web access from the outside, through the satellite feed, set
for our purposes at 64kbps up and 300 Kbps downlink (the lowest
speed option with Tachyon) was quite satisfactory.
http://168.215.212.132
Satellite
to Data Logger Performance - NL100 Interface
The big
step from just using the satellite Internet for bi-directional
web access, or sending receiving email or files, to linking
interactive software to data loggers is another matter.
In the
first place, almost all Campbell and Sutron, as well as Hobo,
Pacom data loggers are designed to be accessed by low speed
serial connections, with 9,600 baud being the most ubiquitous.
The reason was simple. Numerical data collected over time from
sensors attached to the most used data loggers makes pretty
small files. Even months of collection result in data files
no more than a few megabytes of numbers. Since processing that
data - putting it either into a data base, or converting it
into graphical charts - takes place in the computers in research
centers, and not in the data loggers themselves - the flow of
data from the logger to the disk of a computer can be a small
serial line task.
But the
Internet does not run on serial, or circuit switched, connections.
It routinely is based on IP packets and packet switching, and
uses Ethernet 10baseT cabling and protocols. Even though a cell
phone, with a 9,600 baud serial modem connection to a data logger,
can fetch a day's worth of data from a logger in just a minute
or so, the traditional Campbell Scientific software for pulling
data out of a logger or a storage module, only 'talked to' the
serial port on a base computer. PC208W software, up through
Version 3.1 operated this way only. So while serial data radios
- such as Freewaves, or World Wireless Hoppers or Microhoppers
- could be used to link a PC with a remote CR10X or 23X - which
is what we have deployed in both Luquillo and Trout Lake LTER
field sites, neither remote, Internet connected researchers
could access the data logger through the network, nor could
higher speed ethernet-connected radios, such as Aironets, YDI's
or Teletronics be used. Nor could a Tachyon satellite feed be
directly connected to a data logger or set of them - since the
data loggers produce single-channel serial, and the satellite
feed comes out ethernet and IP packetted. Until now.
Campbell
has just started to ship a new version of PC208W software. Version
3.2. With its companion hardware, called NL100. It is a partial
solution to the Internet accessible Data Loggers.' Partial,
because it was not actually designed and priced to be put at
the location of each data logger - which at $500 apiece is rather
stiff when a CR10X costs only $1,000 - but in universities and
offices where the researchers nay be on a network - ethernet
connected. One NL100, which translates an Ethernet/IP connection
to a serial connection can be set up at the research base station,
run PC208W V 3.2, and access the NL100 by IP routing and ethernet,
then come out of the NL100 device by modified RS232 serial (CSI/0)
and thence into a serial radio such as the Freewave to reach
a data logger.
NL100
is in the foreground, cover off. Ethernet in at the right, CS
I/O out to data logger in the rear via serial cable to CS I/O
into the CR23X.
This
solution DOES contain promise for a different problem which
will increase in the future - 'splitting' an IP/Ethernet data
radio signal at a small field 'hub' location so that one branch
can go into the NL100 and thence into a CR10X or 23X, while
the other branch goes into a IP/Ethernet connected device such
as the Axis Web Cam - both being at the end point in the field.
This is similar to what we need to do at the Poker Flat/Caribou
Creek confluence in Alaska. Over one 2Mbps data radio being
able to service both the web camera, with its ethernet access
and IP number inside, and a serial connected 9,600 baud CR21X.
Then
the NL100 also may solve our Satellite IP to data loggers problem.
So, in a bench test mode, Dan Withers from Seattle, lately working
for World Wireless, but now able to work for our NSF project
in various ways, and I got the NL100 and PC208W v3.2 working
first in its intended-design mode.
A. The NL100 hardware
has three ports - a 10baseT Ethernet port, a 9 pin RS232 port,
and a modified RS232 port called 'CS I/O' (which matches the
voltages of the corresponding CS I/O port of the Data Logger,
and thus require only a serial cable between them, rather
than the $175 SC932 DTE/DCE Adapters which are required between
standard serial connectors on devices such as Freewave or
World Wireless radios)
B. Using a serial
cable and Hyperterm running in a Windows PC, pulling up the
simple, sequential-settings menu, in the NL100. Setting an
IP address in the NL100 - 216.7.45.67, and its associated
Subnet Mask number 255.255.255.248 - from our assigned block
associated with the Tachyon/Red5 system.
We left all other menu choices at their 'default'
settings. Which includes a novel 'Serial Server Port Number'
that must be assigned to the NL100 and entered in a corresponding
place in the PC208W software setup.
This required 'Port Number' appears to have been
put in by Campbell in order to differentiate one NL100 from
another one in the same local area network, permitting the
same IP number to be used in all NL100s. This may give us
problems later if we attempt to use anything other than the
NL100 to convert serial to ethernet-IP in a wireless network.
There are several low cost OEM devices from Lemos that can
convert an IP, Ethernet, signal, to serial. The NL100 costs
$500, and has a number of additional capabilities besides
the simple conversion of IP to Serial. With future capabilities
built in, such as SLIP protocol which eventually will work
with later generations of Data Loggers that incorporate IP
addressing right into the logger.
But for our purposes, it is clear, at that price,
that deploying a substantial number of data loggers - like
5, 10, or 20 - each requiring an NL100 at $500, can be rather
costly when the CR10X data logger is only in the $1,000 range.
C. We then configured
the PC208W version 3.2 - which has the additional - to 3.1
or earlier - IP functionality. The ability to add a 'Socket'
device, and assign an IP number and the port number to it.
This version of PC208 software is required for any work with
the NL100 - and is sold as part of the NL100 purchase, complete
with CD ROM and full PC208W manual, with additional sections
on the IP functionality. The NL100 itself has a manual too.
With instructions how to remove the device cover, and use
a jumper to reset to default the whole device, if one screws
up the initial settings and locks oneself out of the device.
Which, of course, unfamiliar with the device, and even the
choice of terms used by Campbell, we managed to do.
D. The PC208W 3.2
software, of course, can still be used with serial devices,
as 3.2 and before were designed to do. So, by default, when
one 'Adds Socket' it is called SOC1 (while Add Device would
generate a Com port name like COM1). One puts the IP address
of the NL100 in it, and the Socket Number, which is this case
defaulted in the NL100 to 6783 in the software.
E. We then set up
a CR23X Data Logger with a serial ribbon cable between the
CS I/O ports on the logger and the NL100. Then we powered
it and the Data Logger - 12 volts with a Marine battery of
the type we would in the field - with the small adapters Campbell
provides. We ALSO, knowing there would be substantial latency
in the eventual Satellite link, we put 1,500 milliseconds
of delay in the PC208W software configurations for SOC1 and
the CR208X screens.
F. Initially, with an Ethernet
cable between the PC and the NL100, and the CS I/O ports
on the NL100 and the CR23X connected, we then ran the PC208W
software, called for the connection, and got the desired
'CONNECT' with the usual date/time of both the Data Logger
and the PC displaying. Which we have learned is the acid
test of whether the PC software is talking correctly to
a data logger. There does not need to be any 'data' in its
memory. The little visual icon of two plugs, either plugged
in 'connected' or separated with wiggly lines emulating
electrical pulses, showing 'attempts' to connect, help inform
the user of what is going on, including - importantly -
when a drop occurs somewhere in the long chain of connections
we are attempting - satellite, wireless, serial.
Satellite
to Data Logger, Directly Connected
The next
step was to test reaching the Data Logger via the Tachyon satellite
link from a PC running PC208W 3.2, but with a hard wire between
the IDU Ethernet port and the NL100.
In other
words, we wanted to see if the latency in the satellite link
makes a big difference in reaching the data logger, without
even the ADDED consideration of additional terrestrial wireless
links.
With
the 'delays' set at 1000 milliseconds or higher in the PC208W
software, I repeatedly, over several days of intermittent testing,
reached the CR23X just fine. And it held the link with only
one drop over about 10 tries.
This
confirms that, even with 800-1500 ms latency in the satellite
link, a reliable connection into the remote field could be expected.
PC
w/PC208W,v3.2 <-Internet-><-Tachyon satellite->IDU<-->NL100<->CR23X
Satellite
to Wireless to Data Logger
But of
course there would be few if any occasions where a satellite
ground station were placed within Ethernet hard wire distance
(2-300 feet) of deployed data loggers. So the final, acid, test
of the Satellite <--> Data Logger link would be to add a pair
of 1-3Mbs ethernet-equipped Part 15 radios between the base
station of the Tachyon Satellite link and its IDU, and the NL100
connected to the Data Logger.
Satellite
IDU<-ethernet>Radio<--wireless-->Radio<->NL100<-serial>CR23X
For this
I used a pair of Teletronics 2.4Ghz, 2Mbps, Access Point radios,
which terminate in ethernet connectors. One to go into the NL100
where it would convert the signal to serial for the CS I/O ports.
The other into the Tachyon IDU Ethernet port (or more precisely,
a Hub, so more than one device could be used) All this was across
the lab room - no long reaches.
'Data
Logger/NL100 connected to a Teletronics Radio Access Point (upper
right), which talks to a second radio that is then attached to
the Satellite Ethernet port.'
Well,
it works. Not as reliably as the setup without the final pair
of radios, but it does connect. The problem seems to be that,
whenever there is a long delay, caused by any number of conditions
- general Internet congestion, or retries for the satellite
link, the 'handshaking' time out latency is exceeded, and the
connection drops, only to reconnect quickly (within 5 seconds).
Of course
'drops' like this during a download of data or upload of compiled
upgrade code, would be unsatisfactory. Over 3 days of intermittent
tests the connection, once established (takes up to 10 seconds)
will hold up from 20 seconds to 3 minutes or more.
However
we have not had the opportunity to 'tweak' the connections,
varying all the possible settable variables - the delays set
in the PC208W software, the radio rates and retries, possibly
the satellite delays. Or even consider a 'buffering' arrangement.
If you
have PC208W v3.2 software running on a network linked to the
Internet, you can try setting a 'SOC1' (or whatever) Socket
Device, with a CR23X Device, pointed to IP 216.7.145.67 and
Port 6783, with 1,500 ms delays, and attempt to connect. (We
will leave it up for a month or more from Oct. 1, 2000). See
if you get a connection.
Implications
We have
demonstrated on a 'Dr Watson come here I need you' level of
experimentation so far, that the model I propose - a satellite
base station placed in a remote area, with independent power
operating it, connected to an array of NL100/Data Loggers linked
by no license spread spectrum wireless radios out to their range
(a few hundred meters to 5-10 miles) each with its own IP number
assigned, can work to permit the fetching of CRX10/21/23X collected
data by any researcher on the Internet running PC208W v3.2 Campbell
Software, and supplied with the requisite IP and Port numbers
for each data logger.
If 10 such
data loggers and NL100 devices were deployed in a remote area,
all connected to the Tachyon base unit by $150 radios, the cost
per month (satellite cost) would work out to be about $80 per
logger per month - with access 24/7 anywhere in the world.
Dave Hughes
Principal
Investigator
NEXT
PREVIOUS