RTTY Terminal Unit References

Over the years, a lot has been written about RTTY terminal unit design, especially with respect to the demodulator section. This list summarizes each of the references. Jim Haynes (W6JVE) has also compiled a pretty complete bibliography.

• Methods of Reducing the Effect of Atmospheric Disturbances, Edwin H. Armstrong, Institute of Radio Engineers, October 5, 1927. Proposes the use of frequency shift keying instead of on/off keying for Morse radio telegraphy. "Atmospheric disturbances" affect both frequencies similarly and are cancelled in the receiver. Includes plots of FSK and on/off signals received at up to 100 wpm.
• The Reduction of Atmospheric Disturbances, John R. Carson, Proc. I.R.E 16,1, p.15; January 1928. Argues that "balancing" of interference as proposed by Armstrong, above, will not work. "In fact, as more and more schemes are analyzed and tested, and as the essential matters of the problem are more clearly perceived, we are unavoidabley forced to the conclusion that static, like the poor, will always be with us."
• Certain Topics in Telegraph Transmission Theory, H. Nyquist, Transactions of the A. I. E. E., pp. 617–644, Feb. 1928. The paper demonstrates that an electronic engineer is a mathematician with a hobby. In general, a square telegraph signal consists of a fundamental at half the baud rate plus harmonics. There is no additional information in the harmonics, so they do not need to be transmitted or received. For on/off keying of a carrier, the bandwidth required is the baud rate / 2 below the carrier to baud rate / 2 above the carrier. Thus the required bandwidth is the same as the baud rate. This applies to the tone filters in AFSK demodulation. Filters for RTTY suggests use of third order filters with the tone filters being 1.2 times the baud rate and the low pass filter 0.6 times the baud rate.
• US Patent 1705211, Radio Telegraph System, L. R. Schmitt, March 12, 1929. Describes a frequency shift radio telegraph system. The transmitter is keyed by contacts adding capacitance or inductance to the oscillator circuit. The receiver uses tone filters. The filters are interesting in that the mark filter is resonant at the mark frequency and has a trap resonant at the space frequency in series with it. The space filter similarly has a trap for the mark frequency. The tone levels are compared in a polar relay with each tone driving one coil. The driving circuits are balanced by adjusting the filament voltage to the relay driver tubes. It appears that the detector tubes are biased class B, so envelope detection takes place using rectification in the relay driver tubes (which conduct only on one half of the tone waveform) and the intertia of the polar relay. The polar relay may also provide some hysteresis.
• Frequency Shift Radiotelegraph, Robert M. Sprague, Press Wireless, Electronics, November 1944.
• US Patent 2443434, R. M. Sprague, June 15, 1948. Automatic Signal Control and Means. Two clampers make two versions of the signal, one clamped with its negative peak at zero volts, the other clamped with its positive peak at zero volts. The average of these two signals is passed on to the slicer. The result is that the peak to peak voltage is centered about zero volts. This appears to mostly be intended to correct for frequency drift in FSK systems instead of for selective fading.
• Let's Look At The Signal!, Don Wiggins, W4EHU, RTTY, Volume 8, Number 6, June 1960. Describes FSK as two AM signals modulated by square waves. Tone bandpass filters should, therefore, be wide enough to pass the sidebands for the first and third harmonic of the keying signal, or about 150 Hz. A low pass filter with a cutoff of about 80 Hz before the keyer (after envelope detection) will reduce the effects of impulse noise. On the transmit side, a single pole LPF with a time constant of about 5 ms is suggested to reduce key clicks.
• US Patent 2,999,925, September 12, 1961. Variable Decision Threshold Computer. Circuitry to adjust mark/space threshold in the presence of selective fading.
• New Horizons in Amateur HF-RTTY Transmission, Jim Haynes, WA6ABF, RTTY, Volume 10, Number 11, November 1962. Extensive discussion. Suggests a low pass filter before the FSK modulator to make the spectrum less "peaky" and more closely match a pre-limiter filter. A limiter improves the SNR when the peak signal exceeds the peak noise, but worsens the SNR otherwise. Following a limiter with a BPF removes instantaneous frequency variations coming out of the limiter. This cleaner signal can drive another limiter which cleans up the signal further. Another approach is to feed back the post limiter filter output back to the limiter input. The loop gain is set such that the circuit oscillates in the absence of signal, but locks on to a signal. Noise introduces instanaeous variations in the frequency of the limiter output. These average out using a wide band FM discriminator, as suggested by Sprague. They may cause tone filters to "dip" due to the instananeous frequency deviation, though this can be limited by the narrow bandwidth of the filters, since it takes time for them to change output. This can further be minimized by the post-detection low pass filter. Article also discusses PLL demodulation.
• Of RTTY - and Filters, Frank VanBrunt, W3TUZ, 73 Magazine, November 1962. Discusses the design of phase linear tone BPFs.
• The W6NRM Radioteleprinter Terminal Unit Mark IV, R. H. Weitbrecht, W6NRM, RTTY, Volume 11, Number 3, March 1963. Circuit has an input BPF, limiter, discriminator, data LPF, and keyer. Hysteresis between the LPF and the keying tube is provided by two NE2 neon lamps in series.
• A New Approach to TU Design Using a Limiterless Two-Tone Method, Frank Gaude, K6IBE, RTTY, Volume 11, Number 6, June 1963. In a typical limiter-based system, there is often a bandpass filter in front of the limiter to reduce noise. However, the filters slow down the rise time and fall time of the tone envelopes. If the tones are equal amplitude, the slower rise and fall times result in a slight shift of all transitions (mark to space and space to mark) to a little later. However, if the tones are not equal level (as during a selective fade), the transitions are delayed unequally resulting in bias distortion. This can be compensated for by adjusting the keying threshold based on the tone envelope level. The "slideback" circuit does this using a combination or diodes, resistors, and capacitors.
• Some Theory and Practice in FSK Converter Design, Don Wiggins, W4EHU, RTTY, Volume 11, Number 3, March 1963. Describes the Alltronics-Howard Model K. This terminal unit uses a limiter and FM discriminator. It is suggested that a "comb filter" before the limiter can be used to reduce noise into the limiter. Other articles here, however, suggest that a sharp filter in front of the limiter yields improper operation with selective fading.
• What Is The Two Tone Detector?, R. H. Weitbrecht, W6NRM, RTTY, Volume 11, Number 9, September 1963. Describes modification of Mark III/IV terminal units to incorporate slideback.
• Two-tone, Shifts and Filters, part 1, Frank Gaude, K6IBE, RTTY, Volume 11, Number 9, September 1963. Two-tone with slideback performs better than FM with limiter. Suggests 45 Hz tone filters (same as bit rate) or 60 Hz to allow for slight mistuning of the receiver. Optimum shift would be where mark is at a peak when space is at a null due to multipath. However, due to varying path lengths, this is probably not practical. Discusses the idea of heterodyning the received tones to another frequency for filtering, similar to the approach used in Dovetron terminal units.
• Two-tone, Shifts and Filters, part 2, Frank Gaude, K6IBE/4, RTTY, Volume 12, Number 2, February 1964.
• A Second Look at Limiterless FSK Detection, Victor D. Poor, K3NIO, RTTY, Volume 12, Number 1, January 1964. A discussion of US Patent 2,999,925, Variable Decision Threshold Computer.
• Filters for RTTY, Victor Poor, K3NIO, RTTY, Volume 12, Number 5, May 1964. Extensive discussion of sidebands and required filter performance. Near optimum is 3 pole Butterworth with bandwidth of 1.2 times the baud rate for the tone BPFs and 0.6 times the baud rate for the data LPF.
• Current RTTY Receiving Techniques, Irvin Hoff and Keith Petersen, RTTY, Nolume 12, Number 12, December 1964. Excellent summary of receiving techniques including FM and two tone AM. Discusses automatic threshold control, tone filter bandwidth, data low pass filters, and more. Includes an extensive bibliography used in making this reference list.
• FSK: Signals and Demodulation, Bob Watson, Watkins-Johnson Company, Volume 7, Number 5, 1980, revised 2001. Summary of filter requirements and automatich threshold adjustment.
• FSK Sidebands, FSK Sidebands, Kok Chen, W7AY, Oct 23, 2005, Kok Chen, W7AY, Oct 23, 2005. Shows FSK spectrum for non-phase-continuous and phase-continuous FSK. Shows spectrum of phase-continuous FSK after passing through a phase-linear bandpass filter. Does not appear to mention the possibility of low pass filtering the data prior to keying the FSK generator.
• RTTY Demodulator Development, Kok Chen, W7AY, September 1, 2011. A good overall discussion of demodulator issues. In general, limiters are bad, and filter bandwidth should be narrow enought to just pick up keying signal fundamental sidebands of each tone.
• Improved Automatic Threshold Correction Methods for FSK, Kok Chen, W7AY, w7ay (at) arrl (dot) net, December 16, 2012. This is the second revision of the article which I first wrote in 2008 (an earlier revision was mostly a cosmetic update). This expanded revision includes references and additional details and diagrams to explain the underlying principles. A discussion on how this ATC pertains to Leonard Kahn's "Ratio Squarer" is included in this revision, together with a comparison of error rates. Section 4 is added to show an improvement over the traditional ATC by applying only the clipper from the optimal ATC.
• Extended Nyquist Filters, Kok Chen, W7AY, February 23, 2012, Expanded: March 7, 2013. In a 1928 paper, H. Nyquist described the necessary and sufficient conditions for a filter to pass a telegraph element without being subjected to intersymbol interference. For a rectangular signaling pulse shape, the Nyquist filter which has the narrowest bandwidth is a Raised Cosine filter. However, in the absence of adjacent channel interference, the wider Matched Filter outperforms the Raised Cosine filter under Additive White Gaussian Noise (AWGN) conditions. This paper describes an effective method for deriving other filters which obey the Nyquist criteria. For a rectangular pulse, these extended Nyquist filters have bandwidths that are in between the bandwidth of the Raised Cosine filter and the bandwidth of the Matched Filter. The method described here can be used to derive Nyquist filters for other signaling pulse shapes.
• RTTY Transmit Filters, Kok Chen, W7AY, w7ay (at) arrl (dot) net, January 14, 2013. We study the effects of transmit filtering on an Amateur RTTY signal (Frequency Shift Keying, with data rate of 45.45 baud and FSK shift of 170 Hz). Specifically, we show (i) the practical limit of how narrow an RTTY transmit filter can be constructed, and (ii) some representative plots of the amount of energy overlap there is between the Mark and Space signals for various filters. The former determines the extent we can help avoid causing interference to an adjacent RTTY station, and the latter can help determine the limit of how narrow a transmit filter can be made when used in conjunction with an AFSK transmitter that has high intermodulation distortion. Prior works with RTTY transmit filtering include a study of what a transmit filter does to a Matched Filter waveform [reference 1] and a study of interference from unfiltered RTTY signals during busy RTTY activity [reference 2]. In section 2, we show how a transmit filter affects the error rates from a demodulator that uses a Raised Cosine filter. When an RTTY signal is filtered, the Mark and Space carriers will overlap temporally. When this overlapping signal is passed through a practical AFSK transmitter, the transmitter's intermodulation distortion (IMD) will cause the spectrum of the filtered signal to widen. In section 3, we will discuss the dependency of the overlap energy on the transmit filter bandwidth. This information can be used to estimate the spectrum widening when the transmit IMD characteristics are known.
• RTTY Filters and the Equalized Raised Cosine, Kok Chen, W7AY, October 23, 2014 (updated: October 29, 2014). The Raised Cosine filter satisfies the Nyquist condition only when the input are impulses; it does not satisfy the Nyquist condition when the input are finite duration rectangular RTTY pulses. Even though some inter-symbol interference exists when the Raised Cosine filter is used to filter rectangular pulses, it is often still preferred to the Matched Filter due to the latter's susceptibility to interference from an adjacent signal. The family of "extended" Raised Cosine strikes a compromise between the wider bandwidth of an ISI-free Matched Filter and the amount of ISI from a Raised Cosine filter. For pulse transmissions, each order of the extended raised cosine filter has successively lower ISI until it converges to a Matched Filter. In this article, we investigate the use of frequency equalization on the Raised Cosine to account for the rectangular pulses, and the use for RTTY demodulation.
• Cascaded BPF Calculator - Enter the number of filter stages, Q of each filter, and frequency. Get overall Q and bandwidth.