The CLOVER waveform and protocol are designed specifically to work in the often hostile HF radio environment. CLOVER dynamically measures the Signal-To-Noise Ratio (SNR), Phase Dispersion (PHS) and error-correction loading (ECC) of every data packet. This data is analyzed and the modulation mode of the very next transmission is adjusted to match the present conditions. CLOVER uses a four-tone multi-level modulation system that includes a combination of frequency, phase, and amplitude shift modulation. To combat multi-path propagation distortion that commonly occurs on HF radio signals, CLOVER uses a very low base data rate - 31.25 symbols per second. By using multiple levels of frequency, phase, and amplitude modulation, the in-packet data rates range from 125 bits-per-second (bps) to 750 bps. The data throughput in ARQ mode is adaptively adjusted from 10 bytes-per-second (byps) to 70 byps as ionosphere conditions very.

Every CLOVER data packet includes Reed-Solomon error correction data coding, the same type of coding used in audio compact disks. This allows CLOVER to correct many data errors without having to repeat a packet. In contrast, AMTOR, PACTOR, and AX.25 packet radio all must repeat data packets to correct for errors. If errors exceed the capacity of the Reed-Solomon encoder, CLOVER will also repeat packets, but only those packets that could not be corrected. We call this our ARQ selective repeat feature and it saves a lot of time. In the highest modulation mode - 16P4A (for 16 levels of phase and 4 levels of amplitude modulation) - 6 blocks of 255 bytes of data are sent in each transmission. If blocks 1, 2, 4, and 6 are received correctly but 3 and 5 have uncorrectable errors, data from blocks 1 and 2 are passed to the PC, blocks 4 and 6 are stored, and only blocks 3 and 5 are retransmitted. Unlike AMTOR, PACTOR, or AX.25, CLOVER does not retransmit data once it has been received correctly. The combination of Reed-Solomon error correction coding and selective ARQ repeat provides very efficient and fast error-free HF data transmission. CLOVER ARQ mode is also bi-directional - there is no OVER command. Unlike AMTOR and PACTOR, either station may transmit data at any time. You don't have to wait for the other station to finish or stick in weird "OVER" commands - just type. Forget OVER, MASTER, SLAVE, ISS, IRS, and all the other "ARQ Alphabet Soup". CLOVER is user-friendly and much like using a full-duplex telephone line modem - but on HF radio.

CLOVER transmissions are "transparent", sending all 8 data bits. Any of your computer files and programs may be sent via CLOVER, error-free and without modification. CLOVER ARQ mode includes code compression for rapid file transfer, often reducing the size by 2:1 or by as much as 10:1, further reducing the time it takes to send a data file. Using extensive DSP signal processing, the CLOVER transmit waveform is specifically designed to occupy the narrowest frequency spectra possible. The CLOVER occupied bandwidth is 500 Hz, down to 50 dB below the peak amplitude. This means that CLOVER signals may be spaced as close as 500 Hz apart with no co- channel interference. In contrast, AMTOR signals must be spaced 1 kHz apart, PACTOR 1.5 kHz apart, and HF packet signals 2 kHz apart to prevent co-channel interference. CLOVER recognizes that present-day high frequency bands are very crowded and that we must all transition to bandied efficient modulation techniques to accommodate more users per kHz of HF spectra.



Fast Data on HF Radio by Ray Petit W7GHM

If you operate RTTY, AMTOR, or packet, you've probably seen a few references on your screen to something called "CLOVER." "What is it and why do I need it?" is the usual reaction. Yes, "clover" (little letters) is a plant, sometimes "wished over" (and sung about by Arthur Godfrey). But "CLOVER" (big letters) is a new way to send data on HF radio that Ray Petit, W7GHM, has invented. This is the story of CLOVER, a project that continues to this date.


CLOVER had its beginnings about 15 years ago when Ray and others were experimenting with very narrow bandwidth Morse code. It's called "Cohernt CW." When packet radio came along, Ray tried packet on VHF and then HF. As most of us have found, HF packet radio leaves a lot to be desired. The ionosphere is just not very kind to packet data, and often many repeats are required to pass any data at all on 20 meters. Unlike the rest of us, Ray quickly realized that putting "bandaids" on HF packed or AMTOR was just not going to do the trick at all; what was needed was a new approach. The new approach must be based on a thorough analysis of the real HF signal conditions and on techniques that can compensate for these conditions.

Ray started by just listening to real radio signals, observing their fades and phase changes on typical HF paths and under varying conditions; shortware broadcast signals made great "test signals." Combining information from signals observations, reading, and previous work on Coherent CW, Ray devised a new way to send data on HF radio which he called "Cloverleaf." In July 1990 Ray published the first paper describing the mode in QEX. The name "Cloverleaf" came from the observation of a scope pattern while watching the data; it was a perfect four-leaf clover. As Ray's work continued, the pretty scope pattern was lost ot technology, but the shortened name "CLOVER" remains.

Other amateurs had also been searching for a way to cure the problems we were having with sending data on HF. Bill Henry, K9GWT, and Jim Tolar, W8KOB, of HAL Communications had also been working on the problem for several years. Ray's paper was like a breath of fresh air to us. Finally someone had taken the pains to start at ground zero and devise a modulation format that would work on HF. Very quickly Ray and HAL teamed up to continue work on this new "CLOVER Modulation."

Unique features of Cloverleaf include:

(1) multi-level phase modulation, not FSK

(2) Use of sequential pulses who state changed only between pulses (not when a carrier is on the air)

(3) Very low base data rates (25 bps); and

(4) Very tightly controlled frequency spectra with no sidebands

Cloverleaf could pass error-corrected data over a typical HF path about two to three times faster than AMTOR or HF packet radio. Unfortunately, Cloverleaf also made extremem demands on the HF radio equipment. Radio frequency accuracy and stability had to be of the order of + or - .1 Hz!

This is way beyond the cappabilities of any currently available commercial radio equipment. Ray also designed a complete transceiver to use Cloverleaf modulation.

At the time when Ray and HAL first teamed up, Ray had already started work to include new DSP (Digital Signal Processing) technology in his Cloverleaf modem. DSP offered many advantages over the basically analog Cloverleaf circuitry, the major ones being

(1) greatly reduced radio stability and accuracy requirements (to + or - 10 Hz), and

(2) faster data throughput (to a nice 750 bits per second).

Ray decided to put "all the eggs in nthe DSP basket." The original "Cloverleaf" modem was renamed "CLOVER-I" and the new DSP version dubbed "CLOVER-II."


To adequately explain why CLOVER is such a breakthrough, we must first breifly review the pluses and minuses of existing HF data modes - RTTY, AMTOR, and HF packet radio.

RTTY of course led the way for automatic reception of characters or data via HF radio. RTTY has been around since the 1940s and is very reliable. The techniques we use today to send and receive RTTY are much the same as those first used. We have better equipment, but use the same FSK modulation and Baudot or ASCII code. RTTY is slow and does not offer any error correction or detection. RTTY speeds of 60 WPM (45 baud) to 100 WPM (75 baud) are common. Increasing the RTTY speads increases the errors; we generally stick to 45 baud.

Amtor evolved from an existing ship-to-shore "radio telex" mode, often called "TOR" or "SITOR" (CCIR 4767 and CCIR 625). AMTOR introduced us to a new type of data link -- "ARQ mode" (ARQ stands for Automatic Repeat Request)

AMTOR characters are coded so that the receiving station can detect an error in each character sent. The sending station sends three characters, turns the transmitter OFF, and listens for a character response from the receiving station. The response is either "all OK, send next three," or "repeat last three characters. Like RTTY, it is also pretty "slow". Under the best conditions, AMTOR can pass data at an equivalent RTTY rate of 50 baud 6.67 characters per second) AMTOR is also limited to the same character set as Baudot - just all capital letter and no ASCII control characters.

CLOVER intends to support the many advantages of AMTOR and HF packed radio and fix the major problems of these modes. The most serious limitation of RTTY, AMTOR, and HF packet is data throughput and how the data is used to modulate the radio signal. The ionosphere is not a friendly medium for data signals. HF signals often arrive at the receiving antenna by many different propagation paths; two or more paths are common. Each signal path has its own time delay, amplitude, and even different center frequency. The receiving antenna does not discriminate; it adds all the signals and passes the composite to the receiver. The amplitudes and phases of the separate AC signals combine algebraically to produce a widely varying reciver input. Deep selectives fades and time-smearing of the data pulse transitions are the usual result.

Once combined at the antenna, the individual path signals are not easily separated. It is usually impossible to compensate for all of these "multipath" effects in the demodulator. A good example of multipath ionosphere distortion is the "selective fading" we hear when listening to music from a shortwave radio station. While annoying when listening to music, the distortionn can be totally destructive to data transmissions.

A major nonrecoverable parameter of HF data is the time at which the data state changes from MARK to SPACE, the data transition time. If we lose this information, the modem can no longer tell when one data pulse ends and the next one begins or if the logic state should be a "1" or a "0." When two signals arrive with different propagation time delays, the composite antenna output signal is "smeared" and the transition times overlap. Measurements by Ray and many others show that we can expect this time overlap from different paths to be as much as 3 to 5 milliseconds (ms). Typical at least one half of each data pulse without distortion determine the MARK or SPACE data state. Therefore, the narrowest data pulse which can be reliably demodulated is on the order of 6 to 10 ms, corresponding to maximum data rates in the range of 100 to 167 baud. Observation shows that the 100 baud limit is more realistic and even it can be too high for satisfactory data transmission at times.

HF Packet radio uses a 300 baud data rate, a pulse width of 3.3 milliseconds. Successful HF packet transmissions are therefore very unlikely if the signal is propagated by multiple paths. HF packed works well only when the operating frequency is close to the Maximum Usable Frequency (MUF)-- when there is only one propogation path. Since this is the exception and not the rule, long-term packet performance on a single fixed frequency is pretty poor, and many repeats may be required to pass any data at all.

HF packet radio, AMTOR, and RTTY all use FSK modulation. One radio frequency is sent for the "1" or MARK pulse state and another for the "0" or space state. The Transmitter carrier frequency is shifted back and forth at the same rate as the data. CLOVER uses different modulation techniques. First, CLOVER shifts the phase and not the frequency of the carrier. Second, more than one bit of data can be sent per phase state. For example, BPSK (binary phase shift keying) has two phase state (0 or 180 degrees) which can be used to represent MARK and SPACE. QPSK (Quadrature PSK) has four phase states (0, 90, 180, 270 degrees). A single phase change in QPSK represents the state of two binary bits of data. Similarly, 8PSK can send the state of 3 bits per phase change and 16PSK can send 4 bits per phase change.

CLOVER also allows use of Amplitude Shift Keying (ASK) in the 8PSK and 16PSK modes. We call these modes "8P2A" (4 data bits per phase/amplitude change) and "16P4A" (6 bits per phase/amplitude change.) Since all changes in phase or amplitude occuure at the fixed base rate of 31.25 BPS (an equivalent pulse width of 32 ms), data errors due to multipath time smearing of data transitions are minimized.

The CLOVER modulation strategy is to always send data at a very slow base modulation rate and to use multi-level changes in phase or amplitude to speed up data flow. One final twist to CLOVER-II is taht there are four separate transmitted pulses, each separated by 125 Hz.

Each of the puolses may be modulation by BPSK through 16PSK plus 8P2A or 16P4A modulation. This further multiplies that effective throuput by a factor of four. Putting it all together, CLOVER can send data at rates from its base data rate (31.25 bps) to 24 times its base rate (750 bps) WOW! It's almost like getting something for nothing! Not quite so. There are still problems to be solved.

PSK modulation itself poses some pretty serious problems. IF we modulate a continous carrier using PSK, the frequency spectrum we get is very bad for HF use, as sidebands are strong and extend over a wide spectra. CLOVER avoids this problem by two techniques:

(1) Each of the four tones is an ON/OFF amplitude pulse and the phase is changed only when the pulse is OFF;

(2) The amplitude waveform of each ON/OFF pulse is carefully shaped to minimize the resulting frequency spectra.

Combined, these techniques produce a composite CLOVER spectra that is only 500 Hz wide down to -60dB. This is one half the radio bandwidth required for AMTOR and one quarter that for HF packet radio at an effective data rate of up to 100 times faster than HF packet radio or AMTOR!

CLOVER also takes a different approach to error correction. AMTOR and packet radio both correct errors by sensing errors at the receiver and then requesting repeat transmissions. When there are errors to be fixed, data throuput is slowed by the time it takes to send the repeats. When conditions are getting poor, packet radio bogs down to sending only repeats and no data at all; AMTOR will slow-down considerably under the same conditions.

CLOVER uses a Reed-Solomon error correction code which allows the receiver to actually fix errors WITHOUT requiring repeat transmissions.

For a moderate number of errors, CLOVER doesn not require repeats and the data continous flowing at the no-error rate. To distinguish between the two schemes, we classify AMTOR and packet radio as "error-detection" protocols and CLOVER as an "error-correction" protocol. In addition, like packet radio, CLOVER includes a CRC (Cyclic Redundancy Check sum) which is used when condtions are very bad and the number of errors exceeds the capacity of the Reed-Solomon error corrector.

CLOVER ARQ mode is also adaptive. As a result of the DSP calculations necessary to detect multi-level PSK and ASK, the CLOVER receiver already has information which can be used to determine the signal-to-noise ratio (S/N), phase dispersion, and time dispersion of the received signal. CLOVER has 8 different modulation modes, 4 different error correction settings, and 4 different data block lengths which can be used- a total of 128 different modulation/code/block combinations!

Using real-time signal analysis, the CLOVER receiver will automatically signal the transmitting state to change modes to match existing ionosphere conditions. When propogation is very good, CLOVER can set itself to the higher speed and data literally "screams" down the path.

When conditions are not so great, the data speed is slowed. As noted earlier, the CLOVER character throuput rate under typical HF conditions is abaout ten times faster than AMTOR or HF packet, However when we get one of those perfect conditions, CLOVER will "shift gears" and pass data at 50 to 100 times the speed of AMTOR of HF packet radio. In all cases, CLOVER automatically changes speeds to give the maximum speed that the ionosphere will allow.


We hear this question often. The answer is YES. The reason lies in the defintion of the CCIR Emission Designtor and how that matched our FCC part 97 Rules and Requaltion.

CLOVER bandwidth is only 500 Hz- no doubt about it! Since the CLOVER modulator generates tones which drive an LSB transmitter, the modulation mode is "J2." One possible point of confusions: While CLOVER does use multiple tones and multiple modulation levels, CLOVER is NOT a multiplex emission; we are sending only one data stream over the air. The full CCIR emission designator for CLOVER is "500HJ2DEN." This all agrees with FCC Part 97 Rules and Regulations.