The Philosophy of Fuzzy Modes

"Man is still the most extraordinary computer of all."
John F. Kennedy, May 1963

The philosophy behind transmitting and receiving
text that is directly human readable.

In recent years, there have been examples of electronic equipment (notably household appliances and industrial controls) which provide digital simulations of the natural analogue world. This technique has been given the buzz-word name 'Fuzzy Logic'. Although there is a whole engineering study in 'soft' control logic, in reality all it means to you and I is that the natural analogue world is offered functionality (by way of inputs or outputs) that, while still digital in nature, has such fine differences between settings that the analogue appearance is maintained.

Nowadays digital modes are everywhere, from GSM cellular phones to digital HDTV and D*Star™ radio. Despite this increasingly digital trend in radio modes, one can still be reasonably sure that:

There is however a small but brave group of modes (we shall call them 'Fuzzy' modes) that are human readable, and do not fit into the categories above. I coined the term 'Fuzzy' in 1998 as a way to describe human readable text and graphics modes, where we explore human interaction with the digital communications world. For example: It is important to recognise that the human brain, aided by the eye, the ear, and other senses, forms a most incredible signal processing system, far in advance of anything man has yet made (hence the quote from JFK at the top of this page). In the context of uncoded text and graphics radio reception, the human computer is the central processing device, and essential to the success of the technique. During reception of Fuzzy modes, the human processing system is used for: The only assistance the mere human needs is demodulation of the radio signal, and presentation of the data for acquisition. While human readable modes require a human to be available to perform the processing, the results are well worth while, as the performance that results, particularly with noisy transmissions, is remarkable. Admittedly human readable modes tend to be slow, since the human processor is expected to perform all the above tasks concurrently in real time!

Now we are at the point where we can define the requirements for a Fuzzy mode. It is important that, for highest human performance, no electronic or software data processing that might alter the results is performed prior to input to the human signal processor. This leads us to a summary of the requirements for a truly fuzzy system.

  • The transmissions must be UNCODED
  • The receiver must not decide WHEN data is present
  • The receiver must not decide WHAT data is present

RTTY, ASCII, PSK31 and other digital modes are CODED transmissions. In other words, each character is represented by a code of data bits for transmission, and decoded on reception. An alphabet code table is used in conversion from keyboard to transmission, and reception to display. This leads to a significant improvement in efficiency, and removes the redundancy involved in any method that attempts to send text or graphics without coding. Unfortunately coded systems can be seriously affected by noise - one bit received in error will cause a completely different character to print. A further problem with coded modes is that the synchronism or identification of the gap between characters or bits is also coded, and an error in a synchronising bit has serious consequences - many following symbols will be interpreted incorrectly until synchronism is reacquired. RTTY and other start-stop modes suffer from loss of synchronism when a start bit is damaged.

By definition, Fuzzy modes are UNCODED. Any sample received in error will cause a slight blemish, but cannot change the character to another character. In other words, visual Fuzzy modes are direct printing modes. In addition, no coded start-stop or synchronism information can be used by a truly Fuzzy mode. Typically, Fuzzy modes are designed to be non-synchronous, or to operate with free-running timing (what we call 'quasi-synchronous'), and so get around any need for synchronism.

Incidentally, from the point of view of Fuzzy modes, Morse can be considered to be UNCODED. This is because (for experienced operators at least) Morse represents a natural language, so is directly human readable and a truly Fuzzy mode!

Digital modes are designed to sample the incoming data in a synchronous manner - typically at the centre of each data bit. Even though some techniques sample several times during the data bit, and then vote on the nature of the data bit, the process is still at the mercy of poor timing. By far the biggest source of timing errors on a received radio signal is the ionosphere. Timing variations due to changing time of arrival in excess of 10ms are not uncommon, representing a change in path length of about 3000 km.

Fuzzy modes sample the incoming data in a non-synchronous way. The equipment presents (makes audible or visible for human processing) what is received when it arrives. In analogue terms this is simple. In digital systems this can best be done by over-sampling, in other words sampling the data far faster than is necessary to define each element. Oversampling is an important technique on HF, where data bits can arrive at different times due to ionospheric propagation effects. No data rate synchronous sampling is used.

Digital modes are characterised by their on-off nature. Data bits are 0 or 1 - there is no room for a "maybe". As a result, "maybe" bits result in errors. Some digital modes (for example PacTOR) use a technique whereby analogue samples are taken and correlated until the checksum is met, and other systems (for example MT63 and MFSK16) use powerful error correction, but the process is still ultimately digital, because the checksum requires a precise digital message to be matched, and the output of the correlator or error decoder is a digital message.

By contrast, Fuzzy mode equipment should not decide what data is present - it should display or present the probability of there being data present. For visual modes this is best done by displaying the probability as a grey-scale value. For audible modes this is done by essentially presenting the probability as loudness, although the ear is attuned to many more properties of the signal - modulation, distortion, phase, harmonic and noise content, and does this naturally by simply listening to the signal straight from the receiver.

By intentionally including the redundancy that is inherent in non-coded transmissions, the cues for Forward Error Correction (redundant transmission, noise rejection, character recognition, context recognition) are provided. Combining over-sampling with integration (visual or mathematical integration, or both), and by using grey-scale probability representation, a sensitive and robust Fuzzy mode results.

Not all the past or present day implementations of Fuzzy modes obey all the rules, and performance suffers as a result. Design of a Fuzzy mode requires careful thought!

Copyright � Murray Greenman 1997-2009. All rights reserved. Contact the author before using any of this material.