Multi-Tone Hellschreiber

In the broadest sense, Hellschreiber is the conversion of images into sounds suitable for transmission over any medium. Mostly the images are text characters ie. a font, but full pictures are also possible. Hellschreiber is often called a ‘fuzzy’ mode as the receive end makes no decisions on the signal received but simply displays an analogue image for human interpretation.

Time Domain Hellschreiber, better know a Feld-Hell, is the classical and more commonly used method of transmission. It was invented by Dr Rudolf Hell nearly 90 years ago and used machines to generate and receive a simple on/off keyed signal at a fixed baud rate. Several modern software implementations exist. Lots of information on Feld-Hell and it’s history may be found on the Net.

Frequency Domain Hellschreiber, also know as Multi-Tone Hell, is best described as ‘Writing in the Waterfall’ It is a close relative of Feld-Hell and the subject of these web pages. A character font or image is converted into sounds by software using DSP techniques and displayed at the receive end as an image on a standard waterfall display.

How does Multi-tone Hell work ?

The objective is to form an image or text character on a standard waterfall display at the receive end. A short tone burst will produce a ‘dot’ on a waterfall so all that is required to form a pixelated image or character in the waterfall is lots of short tone bursts applied either in sequence or in combination spanning a set of closely spaced frequencies. The more closely spaced frequencies used, the higher the resolution.

A right to left horizontal scrolled waterfall is best for text and images (Spectran and Argo) The more common vertically scrolling waterfall displays work fine for images but are awkward for text.

MT Hell is a true fuzzy mode, there are no preset speeds or tone spacing’s, The receive end is manually adjusted to match and resolve the incoming images/text characters using the speed, span, RBW and brightness controls of the waterfall display – rather like tuning into an SSB signal but with four ‘knobs’. This can take a bit of practice especially if the signal is being sent slowly.

Alternatively, the Bandwidth and speed of the transmission may be adjusted to match known waterfall settings (eg. for Web SDR’s or QRSS frame grabbers).

Sensitivity – MT Hell is a good weak signal mode with a similar sensitivity to CW (the original fuzzy mode) when sent at similar speeds (wpm), additionally, being a visual mode it may be slowed as much as required for even greater sensitivity.

How is Multi-tone Hell generated ?

The text character/image to be sent is pixelated, each horizontal row of pixels is assigned a tone frequency, the lowest tone usually being the bottom row, the tone frequency is increased in discrete steps for each higher row. The pixel columns are then scanned from left to right producing N tones per column where N is the number of rows in the image. Mathematically this is an inverse DFT. A 5×7 Dot matrix font using 7 discrete tones is very convenient for text sending.

There are two basic methods of transmitting the 7 tones generated by a 5×7 image scan :-

Sequential Multi-tone (S/MT Hell), The tones in each column are sent one at a time ie. time multiplexed, the displayed image is sloped by one image pixel and has a very ‘dotty’ appearance. The transmission path is not required to be linear.

Combined Multi-tone (C/MT Hell) , The tones in a column are combined and transmitted together, each tone is transmitted for a whole image pixel period but at reduced power. The displayed image is upright and more readable, additionally a grey scale is possible by controlling the tone amplitudes. The transmission path must be linear.

S/MT Hell vs C/MT Hell

The detection threshold of a multi-tone Hellschreiber transmission is determined by the energy in each image pixel, assuming all other factors such as character rate, peak TX power etc. are constant.

Using a 5×7 dot matrix as an example (7 row tones) :-

S/MT – each of the 7 image pixel tones in a column are transmitted sequentially at full power but for 1/7th of the image pixel period (time multiplexed). ie. each pixel receives 1/7 or -8.45 dB of the the full TX power.

C/MT – the 7 pixel tones in a column are combined, such that each tone is limited to 1/7 of the full drive voltage to avoid TX overdrive when all 7 tones are present. Thus each pixel receives 1/49 or -16.9dB of the full TX power but for a whole pixel period.

The simple comparison above suggests S/MT has a 8.45dB power advantage over C/MT Hell, but this is not the whole story. Since a pixel tone is sent for a whole pixel period in C/MT Hell and only 1/7 of a period in S/MT Hell, the Resolution Bandwidth of the waterfall display at the receiver end may be set 7 times narrower for C/MT Hell. Narrowing the RBW by 7 times will increase the S/N by 8.45 dB and fully recovers the apparent power difference.

This is not surprising as energy = power x time, so the two methods theoretically have equal sensitivities.

However being a Fuzzy mode the detection threshold of a Hell image is subject to human factors (pattern recognition). Here C/MT hell shows a 1 or 2dB advantage as the displayed image is inherently ‘clearer’ and therefor more readily recognisable.

Another factor involved is the nature of the transmitters required, S/MT Hell transmits one tone at time which may be transmitted efficiently by non-linear (class C) amplifiers running at saturation, where as C/MT Hell requires linear amplification to keep 3rd order inter-modulation products within acceptable limits. ie. the power must be backed off by at least 6dB, and preferably 10dB, to maintain good linearity.

All factors considered, I would suggest that S/MT has a 6 to 10dB detection advantage over C/MT Hell mainly due to the need to back off the TX power to maintain linearity for C/MT Hell transmission.

Also, S/MT Hell, by virtue of it’s wider RBW requirement at the receive end, will show greater tolerant to doppler spreading.

The ‘Splatter’ Problem

All Hellschreiber systems involve keyed tones, any resulting keying splatter will be clearly visible on the waterfall display and reduces the readability of the characters. This is more severe for S/MT Hell as the tones are switched multiple times per image pixel.

Zero crossing/phase preserving switching will reduce wide band spectral spreading but does little for close-in and inter-pixel splatter. Envelope shaping must be employed but is only fully effective when a linear transmitter is used and a raise cosine envelope (hanning window) is applied to the keyed tone. This will eliminate the splatter and produces a near perfect display of C/MT Hell characters on the waterfall. Spectral purity has a high price – a reduction in average TX power and therefor the detection threshold by 6dB !

PP/MT Hell – an new hybrid

Having identified both the strengths and weaknesses of S/MT & C/MT Hellschreiber above, I sought a solution to satisfy the mutually exclusive ideals of good weak signal performance AND good display quality at higher signal levels, ie. a wide dynamic range.

I believe I have achieved this in creating PolyPhase Multi-tone Hellschreiber ( PP/MT Hell), a hybrid of C/MT and S/MT Hell.

C/MT Hell with a raised cosine tone envelope shaping has excellent quality/purity, but, poor weak signal performance due to the high peak to average ratio of the transmitted signal. So the challenge was in finding a way to improving the Peak to average ratio of a windowed C/MT Hell signal.

A solution was found in polyphasing the raised cosine envelope, ie. to stagger the start points of each raised cosine window evenly over one pixel period. thus each tone’s window peak’s at different times during the pixel, this improves the peak to average by 6dB.

PP/MT Hell has a one pixel slope similar to S/MT Hell but with the spectral purity and readability of an envelope shaped C/MT Hell signal.

PP/MT Hell requires a linear transmitter, like C/MT HELL. At least 6dB’s of power de-rating should be applied to the transmitter to maintain spectral purity. ie. a 100w rated transceiver should be run at 25w peak with no Audio or RF compression applied.

In conclusion

S/MT Hell using zero crossing tone switching

– good weak signal performance
– simple transmitter requirements
– very energy efficient
– poor display quality even at high signal levels
– poor spectral purity

S/MT Hell is well suited for QRP use, especially when sent very slowly (QRSS beacons).

C/MT Hell with raised cosine envelope shaping

– Excellent display quality at higher signal levels
– Excellent spectral purity
– Suitable for high power transmission
– less sensitive (10 to 14dB less than S/MT Hell)
– requires a linear (SSB) transmitter

PP/MT Hell with polyphased raised cosine shaping

– same a C/MT Hell but 6dB more average TX power

PP/MT and envelope shaped C/MT Hell are spectrally very clean and suitable for high power
transmission in the crowded HF bands, both make good ‘chat’ modes at typing speeds (30 wpm).
PP/MT Hell having a 6dB detection advantage as a weak signal mode.

All MT Hell systems have no specific timing requirements, never print wrong characters and work
with standard waterfall displays

Acknowledgements

ZL1BPU Murray Greenman ,  S/MT Hell’s creator and author of the Definitive web site on MT Hell
ZL2AFP Con Wassilieff       , Murray’s colleague, thanks both for the first tests of my MT Hell signal. I2PHD Alberto di Bene       , DSP genius and creator of world class SDR spectrum display software.