Autocorrelation

The following was quoted/adapted from the WUN0509 Digital Review column titled "Autocorrelation - What is it? " by Day Watson pub. Sept.1999 (c) Worldwide Utility News

The basic parameters measured on a typical FSK signal are baudrate and shift. However with the more advanced decoders provision may be made for a further parameter - autocorrelation bit.

This is determined by a module which analyses the incoming 1's/0's which have been recovered from the tones received from the receiver. The software analyses these bits over a period of time (usually over a 512 bit period) looking for a regularly repeating bit pattern.

The module display within HOKA's Code3 and Code30 packages produce fairly close random amplitude spikes across the scale (bits). When an autocorrelation bit has been determined at value x a definative spike will be produced at bit x on the scale (and at multiples of x) and rise above the generally random ones.

It should be noted that the baudrate must first be determined and set into this value before a meaningful result can be achieved. The way the repeatable pattern is determined varies between different modes and the following notes indicates these variants. This shows some complex number crunching is required.

Autocorrelation and Various Modes

• RTTY. Whilst the module is basically designed for use with synchronous data it will provide a value for rtty but rtty provides its own little foibles. The typical value is 15. The module cannot give an indication in other than a whole integer and therefore 15 is representative of two 7.5 bit characters. Only rtty which is asynchronous may use halfbit (in the STOP) elements. Given this characteristic we can take the measured (synchronous) speed and halve it to achieve the async speed. Where there is no halfbit element in the STOP then one is likely to determine an autocorrelation bit of 7, 8, 9, 10 or 11.

• ARQ (Sitor/A). This will show as a value of 45. It is determined from the repetition cycle of the system. The period is 450mS (45 x 10mS bit period).

• Twinplex. This is the same as Sitor/A.

• FEC (Sitor/B). This gives a value of 35. This is due to the interleaving system employed; the repeated character (7 bit) starting 35 bits later than the original.

• ARQ/E. One is likely to see a major spike at 56 and 8 minor spikes. This is because two repeatable patterns can be determined. Each character is 7 bits long but the first bit is either 0 if characters (traffic), or 1 if control (alpha, betas, RQ) (betas if idling). This is the first repeatable characteristic. The second comes from the fact that the data is inverted (in respect of the aggregate) every eighth character. A variant is a major spike at 28 rather than 56; where every fourth character is inverted.

• ARQ/E3. Uses the ITA3 7-bit character set. Only when the character is continuously repeated eg BETA when idling will the minor spikes be indicated. A spike at 28 or 56 indicates the inversion of a character against the aggregate every 4th or 8th character.

• ARQ/M2 (ITU Rec.342). As ARQ/E3 re individuals characters. One channel is erect the other is inverted, but each channel has one character in four inverted hence 2 x 28 gives a displayed value of 56.

• ARQ6-90. Similar to Sitor/A with the same repetition cycle time of 450mS. However as the element time is 5mS (200bd) the autocorrelation bit values is 90.

• Swedish ARQ. Like Sitor/A based on the repetition cycle but this mode has three possibilities depending on the number of characters per block in use at the time hence

    Chars/Block     Cycle time      A/cor Bit
     3               450 mS          45
     9               900 mS          90
    22              1800 mS         180

• Pactor I. The cycle period is 1250 mS being (125 bits at 0mS(100bd)) or (250 bits at 5mS(200bd)) The autocorrelation bit will therefore have a value of 125 or 250 depending on the speed in use at the time of measurement. Pactor uses an 8 bit word in the data structure. If ASCII is in use and,being a 7-bit set the, 8th bit is always 0. This bit is repeated every 8 therefore smaller indications appear every 8 bits on the module display between the 125 or 250 spikes. If Huffman coding is in use and using all 8 bits the pattern will be random and no minor spikes will be generated.

• ARTRAC(Dup-ARQ). A value of 88 is due to the 88 bit period of data and the extended pause period. Normally this would be 44 - station burst of 32 bits plus its subsequent pause of 12 bits (96mS/8mS (bit time at 125 bds). If a second station is alternating on the same channel the module software will see two cycles and calculate accordingly.

• FEC/A. Uses the same alphabet as ARQ/E and the module produces a value of 14 and multiples. This comes from the first bit in the chan 1/Chan 2 bit interleaved string which is invariably a repeatable 0 in the 7 bit traffic characters, (or 1 when idling in control characters). Channel 1 is original data in 7 bits with channel 2 being the interleaved 7 bits carrying the check information of much earlier originals.

• POLISH ARQ. Definately an interesting mode for capturing not 1, not 2 but 3 values. These are 7, 35 and 42.
  • 7 This will be found in the return leg of the system where the bit stream is one of the constant beta character (7 bits in CCIR476 coding). Exceptions will be the occasional gamma (RQ) character which appears when a block repetition is required. With the same character this gives the repeat pattern every 7.
  • 42 In the traffic leg if the traffic is offline encrypt ie 5 fig groups of figures each followed by a space. The space character is the repeat pattern occurring every 6 x 7-bits ie 42.
  • 35 Again in the traffic leg, if the data is normal Polish text the characters are random as far as bit pattern is concerned so no apparent value is flagged. But, if the system is experiencing propagational difficulties and operating frequent block resends, a value of 35 may appears during this sequence. 35 bits means a repeat pattern of 5 characters. Polish Diplo currently uses an 5RC setting ie the resent block consists of the repeat of the last four characters plus a preceding gamma (RQ) character. Frequent repeating of this block will be detected and 35 generated by the module.

• NATO 75/850. This is a typical system where the module sees no repeatable pattern; as one would expect in a pseudo-random bit stream. This is a fully encrypted system (one of many such systems) with no chinks in the armour to allow a cryptoanalyst a means of getting into it. This is different to carrying encrypted traffic on an otherwise recognizable system.

from the WUN0509 Digital Review column by Day Watson pub. Sept.1999 (c) Worldwide Utility News

• Return to DXing Digital Utilities
• Return to Selected Articles from the WUN Archives