Data

19 Communications
It is generally assumed that data communication is a modern used for special applications, for example slow PBSK modes and
aspect of amateur radio, but of course Morse code transmission the MFSK mode JASON (weak signal modes for LF); and WSJT
can also be regarded as a data mode. Recent deregulation of and high speed Hellschreiber (for meteor scatter use).
international and local Morse requirements has done nothing to The late 1990s and early 2000s saw the ascendancy of the
weaken the interest in Morse, and the mode remains one of the high-speed personal computer with sound card as the preferred
most popular on the DX bands. method of signal processing. In addition to the development of
The mode of operation first associated with data communica- new modes that would not have been possible without digital
tions by most amateurs is radio teletype (RTTY). Amateurs in the signal processing, the sound card technique has made digital
UK first began using RTTY on the air in the late 1950s with sur- mode operation easy and inexpensive for beginners. It has also
plus machines such as the Creed 7B. Despite inferior perform- paved the way for a range of new tools, such as the spectrogram,
ance in comparison to that of more recently developed modes, used to detect weak signals and monitor propagation.
RTTY still has its enthusiasts. There are still many RTTY con- Although not exactly a data mode, Slow Scan Television (SSTV)
tests, although now computers are used in place of mechanical technology also advanced significantly during this period, to
machines. become a popular addition to the SSB QSO. Through the use of
the sound card and digital signal processing, inexpensive SSTV
DATA MODE DEVELOPMENTS operation is now widely enjoyed. See the Image Techniques
Digital electronics and then computers allowed technically chapter for more on SSTV.
improved data modes to be developed. The first, developed by The trend in HF data modes is continuing to move away from
Peter Martinez, G3PLX, was called AmTOR (from 'amateur tele- specialised hardware toward general-purpose computers using
type over radio'). Based on the commercial SiTOR system, this digital signal processing. Performance never dreamed possible
was the first amateur data communications mode to make use is becoming a reality, and indications are that before long the
of error-detection and correction techniques, albeit in a very sim- receiver and transmitter as we know them will be completely
ple form. This was to be followed by many other specially absorbed into the computer - the so-called software defined
designed data modes. Commercially developed modems offer- radio [2].
ing protocols such as PacTOR, G-TOR and CLOVER were intro- Towards the end of the 1970s, amateurs in North America
duced, and these continue to provide error-free (but expensive) began experimenting with microprocessor and digital tech-
HF communications for bulletin-board and automated mail sys- niques to generate and process data. A standard based on the
tems; these modes correct errors using two-way transactions. CCITT X.25 public packet switched network protocol was finally
Fig 19.1 shows how modern modes have developed from RTTY. agreed on, and this Amateur X.25 (AX.25) protocol became the
There has also been strong interest in modes that (like RTTY) standard for amateur packet radio. During the 1980s and
offered the amateur operator real-time keyboard to keyboard 1990s the technique grew into a global amateur integrated net-
chatting. From the mid 1990s, a serious effort was made to work offering error corrected communications between any two
replace RTTY with modes designed specifically to take into stations in a network, file transfer, automatic message storage
account the requirements of amateur chat modes, and the and forwarding, and bulletin dissemination. While packet radio
characteristics of HF propagation. The first of these new ‘design- no longer has the following it once had, largely due to the incred-
er’ modes was PSK31, developed by G3PLX. An amazingly fruit- ible popularity of the Internet, useful work is still continuing in
ful period of new mode development followed, based on the PC the area of packet data handling. There is especially strong
sound card, and as one experienced operator interest in international message handling for remote users,
commented:"More modes have been developed in the last few computer networking, and in technology related to telemetry
years than in the previous century!" [1]. During this period, and the tracking of vehicles and other assets. The leaders in this
Hellschreiber was revived, MT63 was developed, and MFSK16 area have been Bob Bruninga, WB4APR, (APRS) [3]; Phil Karn,
was introduced. Other less successful modes have come and KA9Q [4]; Ian Wade , G3NRW (radio network operating systems)
gone, or remain as curiosities. Some newer developments are [5], and the late Roger Barker, G4IDE (UIVIEW and WINPACK) [6].
Packet radio has also been widely used to disseminate hints to
DX operators ("DX spotting").

GETTING STARTED
There is no need to be daunted by the prospect of operating HF
data modes. While some systems require skill and experience,
there are also some very effective computer modes that are
easy to set up, and a pleasure to use. The simplest are
Hellschreiber and PSK31.
HF keyboard mode operation is now a very cost-effective addi-
tion to the ham shack. Not only is the equipment inexpensive, if
an HF rig and computer are available, but also the more effec-
tive modes do not require high power or large antennas for good
DX. All the keyboard to keyboard modes now use sound card
Fig 19.1: Development of HF Digital Modes technology, and there is software for the most popular modes

The Radio Communication Handbook 19.1

19: DATA COMMUNICATIONS

necessary across the transformer to fur-

ther attenuate the audio. If an accessory
socket is used, the values shown may
suffice.
While the receiver cable is shown with
a connector to be directly plugged into
the external speaker socket on the
receiver, with many rigs this will discon-
nect the speaker, which isn't helpful. It is
best in this case to use an adaptor allow-
ing both the PC cable and an external
speaker to be connected.
There is a very good reason for not
using the computer speakers instead of
an external speaker on the transceiver.
Computer speakers receive their audio
from the sound card output, and by con-
necting the LINE IN signal from the radio
to the LINE OUT or SPEAKER OUT and
the speakers, the receiver output signal
Fig 19.2: The simple cables used to connect PC to radio
will also be sent to the microphone input
available for several different platforms - the most popular being of the transceiver. This causes feedback problems, especially if
the ubiquitous PC with some version of Windows™ or LINUX™ VOX is used.
operating system. Some excellent commercial software provides
the widest range of features and operating modes. There is cer- The Importance of Isolation
tainly no performance compromise involved in operating using a The transformers shown in Fig 19.2 provide complete DC isola-
computer rather than a specialised hardware solution. tion between the computer and the radio transceiver. The most
To operate these digital modes, a modern computer with compelling reason to do this is to prevent serious damage to the
sound card is required (computer specification depends on radio and computer. Most power supplies are grounded for safe-
software, but even a 233MHz Pentium will provide a lot of ty reasons. If the power supply cable to the transmitter becomes
pleasure). Some simple cables, easy to make, are also loose, the full 20A transmitter current can pass through the
required, or a commercial interface such as the RIGBLASTER™ microphone circuit, down the cable and through the computer
can be purchased. A conventional HF SSB transceiver is used, sound card to ground via the PC power cable. Even if the trans-
connected via the receiver audio output and microphone audio mitter power cable is considered reliable, significant current
input. It is best if this is a modern solid-state unit, with good fil- could still flow through the sound card cable, causing instability,
ters and low drift, but many operators use older rigs for RTTY, hum and RF feedback. The simple expedient of isolating the
MT63, Hellschreiber and SSTV with no particular problems. connections also reduces the risk of RF in the computer, and
These are the modes least affected by drift and poor frequency computer noises in the radio.
netting.
Some of the newer modes require very high stability, and very VOX and PTT Control
low frequency offset is necessary between transmit and receive. Most operators find VOX operation of digital modes quite appro-
Most synthesised transceivers will suffice. A transceiver that priate and reliable, although the delay may need to be set longer
drifts less than 5Hz per over will operate the newer modes very than for Morse or SSB. If for some reason direct control of the
successfully. Unfortunately offset cannot be accurately correct- rig is necessary, the transmit control must also use an isolated
ed by using the Receiver Incremental Tuning (RIT). circuit. An opto-coupler does this nicely, driving the Press - to -
The connections between computer and transceiver are quite Talk (PTT) directly without requiring a relay or any further power
straightforward. Most amateurs should be able to build the supplies.
required cables. See Fig 19.2. The digital mode software usually controls the transceiver via
The resistors in the transmit cable are used to attenuate the a serial port, by driving RTS or DTR (often both) positive on trans-
sound card signal so that it does not overload the transceiver. If mit, with an appropriate delay before sending tones out from the
the microphone socket is used, a lower value of resistor may be sound card. The design in Fig 19.3 is an appropriate PTT circuit
for a transceiver with positive voltage on the PTT line
and a current when PTT is closed of up to 100mA or
less.
Many transceivers include an ‘accessory socket’,
offering line-level audio inputs and outputs for transmit
and receive. Using these instead of the microphone
socket and speaker socket can be really convenient,
but can lead to a range of unexpected problems.
Sometimes PTT is not available from the accessory
socket, and sometimes the VOX does not operate from
this socket. The signal levels can also be quite different
to the speaker and microphone connections. Even
more troublesome, some transceivers leave the micro-
Fig 19.3: An opto-isolated PTT circuit phone operating while sending data through the acces-

19.2 The Radio Communication Handbook

 19: DATA COMMUNICATIONS

summary of the specification of

the RTTY mode is shown in Table
19.1.
The RTTY technique as we know
it took many decades to become
what it is today. Engineers Emile
Baudot (who developed the system
of multiplexing the data), Donald
Murray (who developed the actual
alphabet), Frederick Creed (who
sold telegram printers to the GPO),
Howard Krum (who developed the
start-stop technique), and finally
Edwin Armstrong (who developed
the FSK keying technique) all
played significant parts.
Fig 19.4: A simple way to connect microphone and computer
Perversely, the alphabet developed
sory socket, so coughs, mutterings and keyboard clatter go out by Murray is also known as the Baudot code, although it wasn't
over the air! developed by Baudot, and is an alphabet, rather than a code!
A simple home-made adaptor (see Fig 19.4) provides a way to The ITA2 five-pulse alphabet has only two conditions for each
operate voice and data modes interchangeably without discon- of the five data pulses (binary 0 or 1, space or mark), allowing
necting anything. Using this design, the data transmit cable is for 32 different combinations. Because it was necessary to pro-
connected by default, but when the microphone PTT switch is vide 26 letters, 10 figures and punctuation marks, the 32 com-
depressed, the relay switches over the audio input and normal binations were not enough, and the problem was resolved by
microphone use occurs. Operation is simple, and feels natural. using most combinations twice; once in letters (LTRS) case, and
The isolated PTT circuit of Fig 19.3 can be built into the same again in figures (FIGS) case. Two special characters were
box, and the whole assembly replaces the transmit cable in Fig assigned, LTRS and FIGS, to indicate which case was in use. The
19.2. There are several similar designs offered as kits [7]. receiving station continued to use the case indicated by the last
There are several commercial interface designs available for received case command character until it received a different
users not disposed to building a home-made or a kitset inter- one. Control functions such as LTRS, FIGS, CR, LF, space and
face, but not all provide full isolation. There are also USB inter- blank were made available in either case. The remaining 26
faces suitable for laptop computers and others with no serial have different meanings depending on whether the LTRS or
port or sound card. FIGS case is selected.
An area that causes confusion among beginners is the busi- Each mode in this chapter is accompanied by a spectrogram
ness of setting up and adjusting the sound card. The adjust- that illustrates the bandwidth properties and appearance of the
ments are all performed in software, mostly using an application signal. Fig 19.5 is the first of these. The spectrogram is a three-
provided with the operating system, and once set for one mode dimensional record of the radio signal - frequency vertically, time
or program, the settings should be correct for all the rest. There
are two main software adjustments, for transmit and for receive,
and it is not very obvious where to find these, especially the
receiver adjustments. The better applications provide direct
access to the adjustments. In addition to the gain settings, you
need to select the correct inputs and outputs, and disable those
not being used. The procedure and these adjustments are
described in detail in the RSGB publication Digital Modes for All
Occasions [8], a reference work recommended for both novice Fig 19.5: The RTTY Spectrogram
and experienced operators.

RTTY
RTTY is now almost exclusively operated using computers, but
the actual data signalling remains the same as it was in the
1950s, when mechanical machines were used.
RTTY uses five sequential pulses to represent each of the let-
ters, figures, symbols and machine functions. Start and stop
pulses are added to facilitate serial transmission. This code is
now recognised as the International Telegraph Alphabet No 2
(ITA2) [9], an international standard with national variations. A

Symbol Rate 45.45 or 50 baud

Typing Speed 60 or 66WPM
Bandwidth 270Hz
ITU-R Description 270HF1B

Table 19.1: RTTY summary Fig 19.6: Block Diagrams of hardware and sound card systems

The Radio Communication Handbook 19.3

19: DATA COMMUNICATIONS

horizontally, and signal strength indicated by brightness. All the

Symbol Rate 100 baud
spectrograms were recorded from live signals at the same
Typing Speed 66WPM
bandwidth and time settings, allowing the bandwidth to be
Bandwidth 400Hz
assessed and should assist the operator to identify the signal
ITU-R Description 400HF1B
from its appearance.
Compatibility was a problem in the early days of RTTY.
Different types of surplus mechanical equipment was being Table 19.2: AmTOR summary
used in different countries, many with different signalling block repetition time is 450ms, so there is 240ms in each cycle
speeds, and the American 45.45 baud speed eventually when the ISS is not transmitting.
became the standard. At 45.45 baud each element is 22ms This 240ms period is taken up by the propagation time
long, and each character takes about 165ms to send, so there between stations, time for the IRS to send back its link informa-
are about six characters per second, or 60WPM. tion, time for the return journey back to the ISS, and an
So that the signals can be carried on a radio transmitter, the allowance for switching delay from transmit to receive. This time
low frequency data pulses must be modulated onto a carrier or should be less than 20ms. The 450ms block repetition cycle lim-
tone (sub-carrier). Similarly, at the receiver, the audio frequency its the distance over which a Mode A QSO can take place.
sounds must be demodulated back into pulses. Equipment to do A spectrogram of an AmTOR Mode A transmission is shown in
this is called a terminal unit (TU) or modem, although in com- Fig 19.7, and one showing AmTor Mode B is in Fig 19.8. Table
puter systems a sound card and software inside the computer 19.2 shows a summary of the AmTOR specification.
now perform these functions (Fig 19.6). Most operators now use When it is required to transmit to no particular station (for
audio frequency shift keying (AFSK) and an SSB transmitter. example when calling CQ, operating in a net, or transmitting a
bulletin) there is no one station to act as IRS. Similarly, Mode A
AMTOR isn't helpful to others listening in, as they do not get corrections,
RTTY suffers from problems such as multi-path reception (fad- but do receive repeats they may not need. Mode B is designed
ing) and noise, which make successful decoding difficult or for these applications, and achieves a simple forward error cor-
impossible on many occasions. In order to overcome these prob- rection (FEC) technique by sending each character twice. In
lems, it is useful to be able to compare multiple versions of the order to provide time diversity, each character is repeated after
same transmission. This can be achieved by using frequency, four other characters have been transmitted, thus avoiding
polarisation, space or time diversity. AmTOR uses time diversity, errors associated with bursts of noise. The receiving station
sending groups of characters twice, spaced by a small time tests for the constant four to three ratio, and prints only correct
interval. AmTOR has two modes, Mode A (ARQ - automatic characters. If neither version is correct then an error symbol is
repeat request) and Mode B (FEC - forward error correction), displayed.
which use this time diversity in different ways. In Mode A, a AMTOR Mode A is little used these days, although commercial
repeat is only sent when requested by the receiving station, traffic is still widely heard. Unfortunately timing restrictions
while in Mode B each character is always sent twice. make PC sound card programs for AMTOR Mode A impractical.
AmTOR was developed from the commercial SiTOR system, AMTOR Mode B has enjoyed a longer life since it can be suc-
which was devised to improve the communication between cessfully operated using a computer with sound card. It is a use-
teleprinters using the ITA2 alphabet. The system uses a seven- ful mode for bulletin broadcast.
pulse alphabet with an exact correspondence to the ITA2 five-
pulse code - the two extra pulses provide error detection infor- PSK31
mation [10]. This Moore code was designed to have a constant Also developed by G3PLX, and based on an idea by Pawel
ratio of four binary 1s to three binary 0s in all valid combina- Jalocha, SP9VRC, PSK31 was intended to replace RTTY as a
tions, so only 35 out of the possible 128 combinations are valid. simple to use and easy to tune keyboard chat mode. The PSK31
This provides a form of error detection since any character which mode first used a low-cost DSP starter kit as modem, but is now
does not have this 4:3 ratio is known to be in error and can be firmly in the realm of PC sound cards and public-domain soft-
rejected. In addition to accommodating the 32 ITA2 combina- ware, using modern DSP techniques [11]. The bandwidth of
tions, a further three are available for link information signals PSK31 is much lower than most other data modes, and has high
(Idle Signal Alpha, Idle Signal Beta and Repeat request (RQ)). sensitivity, which means it can work at lower signal levels in
AMTOR Mode A is a synchronous system, which transmits today's crowded bands.
blocks of three characters from the transmitting or information Keying is achieved by phase-shifting the carrier by 180°,
sending station (ISS) to the receiving or information receiving rather than frequency-shifting it, resulting in a very narrow-band
station (IRS). During a QSO, the roles switch as the direction of signal. The technique is called differential binary phase shift
traffic changes. The ISS sends its message in groups of three keying, or BPSK. Data is encoded in the phase difference, rather
characters, pausing between groups for a reply from the IRS. than absolute phase, since phase is not constant due to ionos-
The signalling rate is 100 baud, with each character accounting pheric effects. With the chosen baud-rate of 31.25, the band-
for 70ms and a three-character block occupying 210ms. The width is down from the 300-500Hz of other modes to only about

Fig 19.7: The AmTOR-A Spectrogram Fig 19.8: The AmTOR-B Spectrogram

19.4 The Radio Communication Handbook

 19: DATA COMMUNICATIONS

Fig 19.9: The PSK31 Spectrogram Fig 19.10: The Feld-Hell Spectrogram

Symbol Rate 21.25 baud Symbol Rate 112.5 baud

Typing Speed ~35WPM Typing Speed 25WPM
Bandwidth 60Hz Bandwidth 350Hz
ITU-R Description 60H0J2B ITU-R Description 350HA1C

Table 19.3: PSK31 summary Table 19.4: Hellschreiber summary

62.5Hz. By using an alphabet with properties similar to Morse, in the 1930s, was an audio sub-carrier mode, and was soon
ie with short codes for common letters, the text speed of PSK31 sent by radio, predating RTTY for this purpose by at least 15
is about 35WPM. By using narrow filters in the receiver, the per- years. The mode was used to send press traffic right up to the
formance of PSK31, even without error correction, is certainly 1960s.
better than RTTY and AMTOR. In addition, the transmitted signal Developed by Rudolf Hell [13], the technique involves sending
is carefully shaped to minimise bandwidth as the phase is each character as a pattern of timed dots, rather like a dot matrix
switched [12]. printer [14]. Black dots are sent (key down) and white spaces are
A spectrogram of a PSK31 transmission is shown in Fig 19.9, not sent (key up) by scanning each character vertically upwards,
and Table 19.3 gives a summary of the specification. then moving along from left to right at a constant rate. The column
PSK31 has no error correction, but on average gives much bet- data can be analogue, but typically consists of 14 black or white
ter reception than RTTY. Radio paths with fading and especially dot positions per column, and seven columns per character,
high phase shift caused by ionospheric Doppler effects can prove including the space between characters. Each character takes
very difficult, as the incidental phase shift can easily exceed the 400ms to send, so the typing speed is about 25WPM. At the
differential phase shift of the intended modulation. A version of receiver, the incoming dots are presented to the reader as dots of
PSK31 with convolutional error correction (QPSK31) works well varying greyness according to signal strength, allowing the reader
on VHF and paths with burst noise, but is even more adversely to discern the transmitted text by eye, and so enabling text to be
affected by Doppler on HF. This is because the phase shift is recognised in the presence of considerable noise.
reduced to 90° in order to accommodate twice as much data A number of clever techniques were developed to improve
(Quadrature PSK). Several other successful variants exist - for reception and minimise transmission bandwidth. For example,
example FSK31, developed by UT2UZ, which uses MSK modula- the received dots are displayed twice, spaced vertically, making
tion, and PSK63F, developed by IZ8BLY, which uses convolution- synchronisation unnecessary. The font developed by Hell did not
al FEC on a single 62.5 baud PSK bitstream (rather than QPSK). permit individual pixels to be sent, rather at least two were sent
PSK31 is popular on the HF bands, and is probably the most consecutively, so the characters could have 14 x 7 resolution
widely used digital mode. It is also effective on VHF. The latest with the bandwidth of a 7 x 7 font. These features are retained
software is easy to use, and little transmitter power is required today in PC sound card software for Hellschreiber. Modern tech-
for good DX. Perhaps the most popular spot is around niques include rendered characters (grey pixels on corners),
14.070MHz on the 20m band. raised cosine dot shaping for minimum bandwidth, and propor-
tional fonts, which are faster to send.
HELLSCHREIBER The most popular Hell mode is the original one used over mil-
There is a grey area (pun intended) between digital data modes itary radio links from 1944 by portable mechanical machines
(such as RTTY and PSK31) and analogue data modes (such as such as the Siemens A2, and for that reason is called Feld-Hell.
SSTV or HFFAX), of modes which cannot adequately be The signal is on/off keyed at 122.5 baud with carefully shaped
described as totally analogue or totally digital. dots, and has a bandwidth of about 350Hz. Feld-Hell is espe-
Arguably, Morse is one of these modes, because while it is cially useful on noisy bands, and because the transmitter duty
apparently a digital transmission, reception occurs at an ana- cycle is only about 20%, is ideally suited to QRP and portable
logue human-readable level. For example, experienced Morse operation. It is badly affected by multi-path, which causes inter-
operators can identify the sender by his ‘fist’, can read the sig- esting ghosting effects. These can often be minimised by care-
nal better than any electronic means, and can tell much about ful adjustment of receiver gain. Another popular mode is FM-
propagation and the transmitter from the sound of the signal. Hell, developed by Nino Porcino IZ8BLY, which uses minimum
Of the other modes in this category, Hellschreiber is the shift keying (MSK), has similar bandwidth, but is more robust
favourite. The term Fuzzy Modes has been coined to describe and sensitive. FM-Hell is not so affected by multi-path, but oper-
these modes with both analogue and digital features. Fuzzy ates the transmitter at 100% duty cycle.
modes use the human brain to assist in interpretation of an ana- A spectrogram of a Feld-Hell signal is shown in Fig 19.10, and
logue presentation of the received signal, rather than electronic Table 19.4 shows a summary of the specification.
decisions made by hardware or computer. There are many free software packages for Hellschreiber
While Hellschreiber is a relatively recent arrival on the ama- modes, the most popular being IZ8BLY Hellschreiber, written by
teur scene, its origins are old. It was developed as a means of Nino Porcino IZ8BLY. This software includes several other inter-
sending press messages by telephone line. Hellschreiber, even esting Hell-related modes. Hell signals can be found on most

The Radio Communication Handbook 19.5

19: DATA COMMUNICATIONS

Fig 19.11: The MT63 Spectrogram Fig 19.12: The MFSK16 Spectrogram

Symbol Rate 10 baud Symbol Rate 15.625 baud

Typing Speed 100WPM Typing Speed 40WPM
Bandwidth 1000Hz Bandwidth 316Hz
ITU-R Description 1K00J2DEN ITU-R Description 316HF1B

Table 19.5: MT63 summary Table 19.6: MSK16 summary

bands, and is most popular on 80m and 20m. Check around dard teleprinters. Over the years a number of versions of Piccolo
14.075MHz. with differing numbers of tones and different speeds were devel-
oped, and all exhibited good sensitivity and considerable supe-
MT63 riority to teletype on long distance circuits. These systems oper-
This remarkable mode has been likened to a juggernaut driving ated without FEC. It made sense therefore to develop an
down the high street at rush hour - nobody dares get in its way! improved MFSK mode for amateur use.
Developed by Pawel Jalocha, SP9VRC, this is definitely not a The benefits of MFSK are:
mode to be used on a crowded band, but it has some very spe- • Sensitivity improves with the number of tones used.
cial properties. Not unlike many PSK transmissions at the same • More data can be sent as the number of tones increases.
time, MT63 uses 64 carriers, spaced 15.625Hz apart, each one • Since the same data can be sent at a lower keying rate,
phase modulated at 10 baud. The resulting signal bandwidth is immunity to multi-path effects is improved.
1kHz, and the signal sounds just like noise [15]. • Immunity to interference depends on the signalling rate,
What makes the mode particularly unusual is the way the not the overall signal bandwidth, improving sensitivity and
data is coded. The raw data rate is 640 bits per second, but a robustness.
very strong FEC system is used, and the resulting text rate is These benefits mean that a very robust typing-speed mode
100WPM. The FEC system uses a Walsh-Hadamard transform, can be developed that has high sensitivity. Multi-path reception
where the seven ASCII text data bits index a table of carefully causes timing errors which cause individual keying elements to
selected 64-bit words to be transmitted. These 64 data bits are run into each other. Because the keying rate of MFSK is much
then spread across the 64 tones, and also spread over six sec- lower for the same data rate, the technique is useful for avoid-
onds of transmission. The result is a sensitive mode of incredi- ing the timing problems that cause such difficulty to other
ble robustness, reasonably immune to burst noise and interfer- modes. One disadvantage of MFSK is that it requires rather
ence, and also able to operate under conditions of ionospheric accurate tuning and stable equipment.
instability that would stop most other modes. The first and most successful amateur MFSK mode is
MFSK16 [16], designed by Murray Greenman ZL1BPU, specifi-
A spectrogram of MT63 is shown in Fig 19.11, and Table 19.5
cally for long-path keyboard to keyboard operation. It has also
shows a summary of the specification.
proved to be excellent on 80m where NVIS multi-path problems
Because of its bandwidth and slow turnaround (12 seconds
are extreme. The first MFSK16 computer program was STREAM
between overs!), MT63 is little used for DXing. However, little
by Nino Porcino, IZ8BLY, and it is still the most popular. MFSK16
transmitter power is needed and tolerance to drift and mistun-
uses 16 tones spaced 15.625Hz apart, and a signalling rate of
ing is about 50Hz. MT63 is a good choice for maintaining regu-
15.625 baud. Because each tone represents four bits of data,
lar contact with friends over trans-polar and long path routes.
the data rate is 62.5BPS. MFSK16 uses a very powerful convo-
SP9VRC has recently developed an MFSK mode of similar band-
lutional code FEC system, with an interleaver to provide time
width which uses the same FEC system. This new mode, named
diversity. Both the FEC code and interleaver are tied to the data
OLIVIA, is one of the most sensitive modes designed yet, and
bit weighting, and so no synchronism is required. MFSK16 also
has a typing speed of about 17.5WPM. Both modes have wider
uses a variable length character set like Morse and PSK31,
and narrower variants, and both can be found regularly just
which results in a typing speed of over 40WPM.
above 14.1MHz.
Fig 19.12 shows the spectrum and Table 19.6 gives a sum-
mary of the specification.
MFSK16 MFSK16 is one of the best DX modes, requiring little power,
Multi-frequency shift keying (MFSK) operates like RTTY, but and capable of operating under rather poor conditions. Users
instead of just two tones, four or more are used, allowing more report reception with no errors even when the signal cannot be
data to be sent at a lower keying rate. MFSK has been used heard! The FEC system ensures that copy is virtually perfect until
commercially since the 1950s when the Coquelet electro- printing simply stops as the signal is lost, which cannot be said
mechanical system was developed in Belgium. A more sophisti- for most other modes. There are numerous computer programs
cated electronic system called Piccolo was developed for the offering MFSK16. Most operation is close to where RTTY is
British Foreign and Commonwealth office and publicly demon- found, for example just below 14.080MHz on 20m.
strated in 1963. These two systems were initially designed for The sound of MFSK16 is distinctive, and although tuning the
the ITA2 alphabet, and converted the signals to and from stan- signal is tricky, users soon learn to align the receiver with the

19.6 The Radio Communication Handbook

 19: DATA COMMUNICATIONS

lowest (idle) tone. This tone appears at the start of each over
and during pauses in the transmission (see Fig 19.12).
Other amateur-developed MFSK modes include: MFSK8, also
by IZ8BLY; THROB by Lionel Sear, G3PPT; OLIVIA by SP9VRC;
FSK441 by Joe Taylor, K1JT; Domino by Con Wassilief, ZL2AFP;
and JASON, a narrow band LF mode by Alberto deBene, I2PHD.
FSK441 is a high speed four-tone mode for meteor scatter use.
JASON and Domino use Incremental Frequency Keying (IFK),
encoding the data as differences in frequency, rather than Fig 19.14: The CLOVERII Spectrogram
absolute frequency.
MFSK8 is the same bandwidth as MFSK16, but uses 32 tones Symbol Rate 31.25 baud
spaced 8Hz apart at 8 baud. It is extremely difficult to tune accu- Typing Speed 30 - 500WPM
rately. THROB uses an unusual combination of single and dual Bandwidth 500Hz
tones to encode a restricted character set, and operates at 1, 2 ITU-R Description 500HJ2DEN
or 4 baud. Despite the very low signalling rate, the typing speed
is reasonable, since each signal is a complete character. There
Table 19.8: CLOVERII summary
is no FEC.
Domino is designed for HF band chatting, and encodes each pensive) modems, which operate the specialised modes
character of a limited (6-bit) character set into two successive PACTOR, PACTOR2, PACTOR3, CLOVER II and G-TOR.
tones. The tones are in two interleaved sets of eight, odd and The original PACTOR mode is FSK, not unlike AmTOR, except
even, and the data is recovered by measuring the distance that the data is ASCII, transmitted in longer blocks (1.25s peri-
between successive tones. As you can imagine, if one measure- od) and much better error detection is used. In addition, a
ment is in error, the next will be in error in the opposite direction scheme known as Memory ARQ allows data to be corrected by
- this is the main flaw with IFK. The receiver synchronises easily processing multiple corrupted versions of the same data.
because of the odd-even tone sets, and the order of the tone Compression techniques are used to reduce the number of bits
sets is determined by analysis of the received data. Domino has transmitted.
no error correction in its experimental form, and yet it is remark- A summary of the PACTOR specification is in Table 19.13 and
ably robust and forgiving. The ZL2AFP software is very easy to a spectrogram is in Fig 19.13.
use. Later versions PACTOR2 and PACTOR3 use PSK modulation on
The particular advantage of the IFK technique is much multiple carriers, and are considerably faster and more robust.
reduced sensitivity to drift and poor tuning. For example, The calling and linking functions retain the original PACTOR FSK
Domino can be received while the receiver is slowly tuned modulation mode for compatibility. It is important to appreciate
across the signal! Other similar modes with IFK coding and FEC that these are commercial and proprietary modes (not public
are likely to be developed in the future. domain) and therefore their amateur use may be prohibited or
restricted in some countries. A special hardware modem is
HF ARQ MODES required, and it is also not possible to "listen in" to a transmis-
These modes were developed to provide improved automatic sion in these modes.
operation on HF. When forwarding mail or in communication CLOVER II has a wide range of different modulation schemes,
with a bulletin-board system, it is important for communications but is best described as an orthogonal frequency division multi-
to be letter perfect, or the commands could be misinterpreted or plex (OFDM) system. There are four tone frequencies, each
data corrupted. Since these operations are invariably station to amplitude and phase modulated. Special hardware is required,
station and automated (rather than nets or broadcasts), an ARQ and the equipment can automatically switch between the avail-
mode is more appropriate. able modes in an attempt to provide best throughput. Clover
The first automated systems used AMTOR, which maintained uses Reed-Solomon FEC in addition to its ARQ system. The pro-
links well, but data rate was poor. Some also used HF (300 tocol is proprietary. It is not now used very widely.
baud) packet, which performed very poorly unless propagation Fig 19.14 shows a spectrogram of CLOVERII, and a summary
was perfect. Most systems now use commercial (and not inex- of the specification is in Table 19.8.
Of the ARQ modes, G-TOR is the most similar to AMTOR. It has
the same FSK modulation, but differs in using the ASCII charac-
ter set, and in the use of a very strong Golay FEC error correction
system, which transmits two differently coded versions of the
data. Requests for repeat are reduced because the system is
often able to reconstruct the data from the first transmission,
and if the second is required the ability to reconstruct the data
accurately is enhanced further. G-TOR is proprietary and only
available using suitably equipped hardware. Although a good
Fig 19.13: The PACTOR Spectrogram system, unfortunately G-TOR has never enjoyed wide popularity.

Symbol Rate 100 or 200 baud PACKET RADIO

Typing Speed 66 wpm (300WPM PACTOR 2)
Packet radio was the first true amateur digital, as opposed to
Bandwidth 500-600Hz
analogue, transmission system. This makes the relaying of sig-
ITU-R Description 600HF1B (500HG1B PACTOR2)
nals much more efficient since the data is reconstituted at each
stage of the link and any end-to-end noise and distortion is sim-
Table 19.7: PACTOR summary ply that of the digitising process and not the transmission of the

The Radio Communication Handbook 19.7

19: DATA COMMUNICATIONS

AX.25 Level 2 Link Layer Protocol

Version 2 of the AX.25 Level 2 protocol was adopted by the ARRL
back in October 1984. This protocol follows that of CCITT
Recommendation X.25 except that the address field has been
extended to accommodate amateur callsigns, and an
Unnumbered Information (UI) frame has been added. This pro-
tocol formally specifies the format of a packet radio frame and
the action a station must take when it transmits or receives such
a frame.
Fig 19.15: Block diagram of a typical packet radio station At this link layer, data is sent in blocks called frames. As well
as carrying data, each frame carries addressing, error checking
digital information. One of the other main benefits of this mode
and control information. The addressing information carries
of operation was always assumed to be that the channel could
details of the station which sent the frame, who it is intended for
be shared by many users. Unfortunately the radio-based sys-
and which station should relay it. This forms the basis of many
tems are different to computer networks in that not all stations
stations sharing the channel since any station can be set up to
can receive each other, thus making it more difficult for channel
monitor all frames on the channel, through various stages to
sharing.
monitor only those intended for it and ignore any others. The
As with other methods of data communications, packet radio
error-checking information allows the intended recipient to
commonly makes use of a terminal unit (Terminal Node
determine if the frame has been received free of errors. If this is
Controller, or TNC), either a stand-alone unit or as part of a PC-
the case and the two stations have previously established a con-
based system using the sound card as an interface.
nection, an acknowledgement is generated by the receiving sta-
Very simply, the function of the TNC is to take the arriving data
tion. If errors are detected the frame is ignored and some time
and assemble it into packets which are then passed to the on-
later the sending station resends the frame.
board modem (or PC sound card under PC control) for conver-
sion into audio tones. The receive side of the TNC performs the AX.25 Format
reverse of the tasks outlined. On VHF the transmission speed for
Packet radio transmissions are sent in frames with each frame
most end-user access is 1200 baud with tone frequencies of
divided into fields. Each frame consists of a start flag, address
1200Hz (mark) and 2200Hz (space), with 300 baud and 200Hz
field, control field, network protocol identifier, information
shift being employed for HF applications. These standards coin-
field, frame check sum (FCS), and an end flag. Fig 19.16
cide with Bell 202 and 103 modems for VHF and HF respective-
shows the format of a frame and Fig 19.17 shows a typical
ly. 9600 baud is commonly used for inter-site linking and satel-
address field.
lite communications on VHF and UHF.
A block diagram of a typical packet station is shown in Fig Flag field
19.15. Although the drawing shows a computer, a simple dumb Each frame starts and ends with a flag which has a particular bit
terminal can be used; however, to make use of the full facilities pattern: 01111110. This pattern appears only at the beginning
for file transfer etc a computer is essential. and end of frames. If five 1 bits show up elsewhere in the frame,
a procedure called zero insertion (more commonly called bit
Channel Access stuffing) takes place and a 0 is inserted by the sending station
The basis of a packet radio con-
tact is that each station transmits
some information and receives an
acknowledgement. If no acknowl-
edgement is received then the
information is retransmitted. One
of the main causes of non-receipt
of acknowledgement is collision
with another transmission of
either the main transmission or
the acknowledgement.
Early packet radio experiments Fig 19.16: Format of a frame
made use of a channel access
system in which a station trans-
mitted without checking if the
channel was free. If the transmis-
sion was not acknowledged within
the correct time slot, the TNC wait-
ed a random length of time before
retrying. Current packet systems
make use of data carrier detect
(DCD) - they listen for an empty
channel before transmitting. This
is not a guarantee against colli-
sions, because two stations may
‘decide’ to transmit at the same
time, but it is an improvement. Fig 19.17: A typical address field

19.8 The Radio Communication Handbook

 19: DATA COMMUNICATIONS

and deleted by the receiving station. The receiver will therefore When in the connected state either station may request a dis-
delete any 0 bit which follows five consecutive 1 bits that occur connection which occurs after an acknowledgement is received
between the flag fields. or if no response is received after several attempts.
Address field Packet Operation
The address field consists of the destination field, source field
and up to eight optional relay or digipeat stations. These fields Packet operation currently makes use of the HF, VHF, UHF and
usually contain callsigns and space is available for up to six SHF parts of the spectrum with both terrestrial and satellite links
characters per callsign with a seventh available as a secondary being utilised. In the early days, much packet operation was
station identifier (SSID). This allows up to 16 different packet real-time person-to-person operation, either direct or through a
radio stations to operate with one callsign. The default is an digipeater.
SSID of 0. For example, GM4AUP-0 could be the real-time sta- Most TNCs are capable of digipeat operation and this enables
tion, GM4AUP-2 could be a personal message system (PMS) and stations who cannot contact each other direct to do so by the on-
GM4AUP-4 could be a node station. The SSID byte in the digi- frequency retransmission of the digipeater. As packet became
peater address also contains information as to whether it is more popular the real-time operation tended to be replaced ter-
repeating a frame or not. restrially by store-and-forward systems such as nodes, although
there are Earth-orbiting digipeaters placed into operation peri-
Control field odically from amateur-radio equipped space stations.
The control field is used to identify the type of frame being trans-
mitted and the frame number. Digipeaters
Protocol identifier field Most TNCs can be used as a digipeater as this function is usu-
This field is contained within the information field and identifies ally contained within the AX.25 Level 2 firmware.
what, if any, network layer protocol is being used. Fig 19.18 shows how two stations A and D can connect to
each other using digipeaters B and C. In order for information to
Information field
be passed from station A to station D via the digipeaters B and
The information field contains the data to be transmitted and
C, the information frame must be received by station D and the
can contain any number of bytes, up to a maximum of 256, of
acknowledgement frame received by station A before a frame
information.
can be said to be successfully sent. Digipeaters B and C play no
Frame checksum field part in the acknowledgement process; they merely retransmit
The FCS is a 16-bit number calculated by the sender. On receipt any frames that contain their callsigns in the digipeat portion of
of a frame the receiving station calculates a FCS and compares the address field. If the acknowledgement is not received by sta-
it with that received in the FCS field. If the two match then the tion A then the frame is retried over the whole path. The use of
receiving station acknowledges the frame. digipeaters has reduced dramatically in recent years with the
advent of the network nodes.
AX.25 operation
As previously described, the TNC is the device which assembles Network Nodes
the data into frames as above. When first powered up, the TNC The network node significantly improved the packet radio sys-
is in a disconnected state and is monitoring traffic on the appro- tem as a means of communicating between packet-equipped
priate radio channel. stations in both real time and by the use of mailboxes. The major
In order to communicate with another station it is necessary advantage of a network node over a digipeater is that any frame
to enter the connected state. This is done by issuing a connect which is being transmitted is separately acknowledged between
frame which contains the callsign it is requesting connect status each individual element rather along the whole chain.
with as the addressee. If the other station is on the air it Fig 19.19 shows a system with station A trying to communi-
responds with an acknowledgement frame and the stations cate with station D via the nodes B and C. In trying to commu-
become connected. If no acknowledge frame is received the nicate with each other the information is sent from station A to
requesting station re-issues the command a pre-determined node B and acknowledged back to station A. Node B then pass-
time later and continues to do so until a preset number of tries es the frame on to node C and receives an acknowledgement
has taken place. If no connection is established the requesting back. Node C then passes the frame to station D who acknowl-
TNC issues a failure notification. edges it back to Node C. If anywhere in the path no acknowl-
Once a link is established the TNCs enter the connected or edgement is received then the frame is retried only over the
information transfer state and exchange information and super- part of the path for which no acknowledgement has been
visory frames. The control field contains information about the received.
number of the frame being sent and the number of the last one There are two types of network protocol in use - virtual-circuit
received (0 to 7). This allows both TNCs to know the current link and datagram. In the virtual-circuit protocol the appearance of a
status and which to repeat if necessary.

Fig 19.18: How two stations can connect to each other using Fig 19.19: How two stations can connect to each other using
digipeaters network nodes

The Radio Communication Handbook 19.9

19: DATA COMMUNICATIONS

direct connection between the two stations is provided. In order in use by all major DXpeditions. Some stations remain connect-
to establish communications a 'call set-up' packet is sent ed to the Cluster all the time and have an audio warning from
through the network to make a path to the other station. Once their PC to tell them about any DX that might be available.
this path is established information is sent through the circuit. Internet-based DX Clusters are commonly used as an alternative
Any packets sent do not have the full address of the required to radio-based DX Cluster connections.
path because the network attempts to maintain this path for the
duration of the contact. After the contact is completed the virtu- Satellite Communications
al circuit is cleared by removing the information on the path Using satellites for communications can be a very satisfying
along the network. An example of a virtual-circuit protocol is the achievement. There are several data satellites orbiting the
RATS Open System Environment (ROSE) developed by the Radio Earth. Using a dedicated set-up, it is possible to run an auto-
Amateur Telecommunications Society (RATS) of New Jersey. matic station to track, send and receive mail to several different
ROSE is a firmware replacement for TNC2 clones. The virtual-cir- satellites. Full duplex mode is used, and 9600 baud is the stan-
cuit protocol is not very common in the UK and most networking dard used. Packet signals have even been bounced off the
is done using the datagram protocol. Moon, although the distortion on that path prevents regular
In the datagram protocol each packet contains full network communication.
addressing and routing information. This enables a packet to More information can be found in the chapter on satellite
reach its destination via any route still open, regardless of how communication.
reliable the network may be. The network overhead is greater in
this protocol but it has much greater flexibility and the end user Packet Radio Bibilography
does not need to know the route, only the node nearest him and Further information can found from the following books and peri-
the node nearest the station with which he desires to connect. odicals:
Datagram protocols used in the UK are NET/ROM (and clones Your First Packet Station, Steve Jelly, G0WSJ, RSGB,
such as TheNET), TheNODE and Internet. 1996.
Packet Radio Primer, Dave Coomber, G8UYZ, and Martyn
TCP/IP Croft, G8NZU, RSGB, 2nd edition 1995.
The Internet protocol software was written by Phil Karn, KA9Q, NOSIntro, Ian Wade, G3NRW.
and is more commonly known as TCP/IP which is an acronym for AX.25 Link Layer Protocol, ARRL.
two protocols, the Internet Protocol (IP) and the Transmission
ARRL Handbook, ARRL.
Control Protocol (TCP).
In reality KA9Q's TCP/IP consists of a suite of individual pro- RadCom, the RSGB members’ magazine.
tocols, Address Resolution Protocol (ARP), File Transfer Protocol Information on licensing and policy matters with regard to
(FTP), Serial Line Transfer Protocol (SLIP), Simple Mail Transfer data communications is available from the chairman of the Data
Protocol (SMTP), Telnet Protocol, User Datagram Protocol (UDP) Communications Committee c/o RSGB, Lambda House, Potters
as well as TCP and IP. Bar, Hertfordshire EN6 3JE.
Each station using TCP/IP is a network node with a unique IP
address that has been assigned by the local IP address co-ordi- REFERENCES
nator. The amateur TCP/IP network has been assigned the net- [1] Comment made in conversation by Dr Gary Bold ZL1AN
work name AMPRNET and all amateur addresses commence [2] See http://www.sdradio.org/
with the two digits 44, followed by three digits indicating the [3] See http://web.usna.navy.mil/~bruninga/aprs.html
country code (as an example of a full address '44.131.5.2' is [4] See http://www.ka9q.net/
assigned to G3NRW). TCP/IP is becoming very popular in the UK [5] NOSIntro: TCP/IP over Packet Radio, Ian Wade, G3NRW,
and is said to offer many advantages over 'ordinary' AX.25. TAPR
[6] See http://www.ui-view.org/ and http://www.winpack.org.uk/
DX Clusters [7] The BARTG publish designs from time to time. See
A DX Cluster provides information on DX stations being http://www.bartg.demon.co.uk/
worked/heard along with information on QSL managers, WWV [8] Appendix D, Digital Modes for All Occasions, Murray
propagation and prefixes for example. The operation is not dis- Greenman ZL1BPU, RSGB
similar to mailbox operation but users stay connected to their [9] A table of the ITA2 character set is to be found in Digital
local DX Cluster for as long as they wish to receive announce- Modes for All Occasions, Murray Greenman ZL1BPU,
ments. The type of announcement the user receives is cus- RSGB
tomised to suit his own needs and can be used to select prefix [10] A table of this character set is to be found in Appendix B,
information, band information, mode information or a combina- Digital Modes for All Occasions, Murray Greenman
tion of all those. ZL1BPU, RSGB .
Each cluster is generally referred to as a cluster node and [11] See http://www.aintel.bi.ehu.es/psk31.html
these nodes can be connected together to each other via the [12] The technical details of PSK31 are explained in detail on
packet network. This enables an item of DX information (com- the PSK31 web site.
monly referred to as a spot) to propagated to all other cluster [13] Patent "Device for the electric transmission of written
nodes in the network, thereby in theory enabling all connected characters" received 1929.
users to see this spot in a short timeframe. [14] See www.qsl.net/zl1bpu/FUZZY/Contents.html
DX Cluster spots are quite common on virtually all contesting [15] More details can be found at http://www.qsl.net/zl1bpu/MT63
software, in lots of major SSB and CW contests, as well as being [16] See www.qsl.net/zl1bpu/MFSK

19.10 The Radio Communication Handbook