JT: you hate it or you love it…or?

By   6 December 2016 21:37

In my humble opinion, Joe Taylor, K1JT, is a true innovator for the amateur service.

As a scientist, Joe received several notable awards. The amateur community should be grateful for his efforts and resulting advancement. Many amateurs may forget that our use of frequency spectrum is justified by experiments, with possible spin-off that benefits everyone.

Joe applied a number of fundamental principles to software, that is better known as WSJT (Weak Signal Joe Taylor). It is definitely worth having a look at the references mentioned on his internet pages, they explain a lot.

Joe wrote an article in QST in 2001 explaining how the idea was converted into an application to improve meteor scatter communication. Over time, the software broadened with a number of modes, for various weak signals. The popularity of the software has grown substantially, also because of many additions. On the HF bands, the popularity of JT65 is immense.

One of the distinct benefits of JT modes is the ability to decode signals with very low signal to noise ratios (SNR). The software estimates the SNR relative to SSB (2.5 kHz). Most slow modes decode to a range of roughly -15 to -25 dB. The narrowest (and slowest) modes require even less SNR.

The literature has a lot of information, so I will not repeat it here (which would be no information according to the information theory anyway…).

CW is regarded as the prevailing analogue “weak signal” mode, with a requirement of about -10 dB SNR to be decoded. From my experience and thus practically speaking, JT65 contacts can be made without too much difficulty with SNR’s down to about -20 dB. So, JT65 has roughly 10 dB “headroom” when compared with CW.

JT modes not only allow contacts to be made with lower SNR’s, but automatic decoding adds several distinct possibilities that analogue modes do not offer (one should note that CW is of a digital nature, which therefore can be decoded automatically, but with inherent limited reliability, because it lacks error correction).

Automatic decoding
Opposed to CW and older digital modes, most JT transmissions contain extra bits to facilitate error correction. This not only helps to overcome errors, caused by interference for example, but it also results in higher reliability. The resulting probability of erratic decodes is very low.

Automatic decoding leads to a number of features, just to name a few:

  • Saving messages or logging QSO’s
  • Actions like clicking on a message to start a call, advance with the QSO sequence, synchronise frequency etc.
  • Reporting received messages via internet (a.o. pskreporter.info, hamspots.net, wsprnet.org etc.)
  • Plotting locators on a map
  • Recognising your own call (highlight a message when someone calls you)
  • Automatic sequencing
  • Search for calls, regions, of other elements of interest

It is also noteworthy that the entire audio passband can be decoded, so you basically skim a frequency band. It also enables split frequency operation within the passband. If you expect many callers to call a DX station, it can be useful not to call simplex but to shift the transmit frequency to minimise interference from other callers, increasing the likelihood of you being answered.

Propagation research
The WSPR (whisper) mode is recommended for beacon transmissions. It is very slow (WSPR-2 messages take nearly 2 minutes) which results in a very low SNR threshold. Transmissions with milliwatts get across incredible distances. The reports (or spots) can be uploaded to wsprnet.org where one can create queries to filter out bands for example.

At the time of writing, I am working on an application to analyse propagation, using data from wsprnet.org.

Another current project is to examine the propagation between my QTH and VK7BO on 60 metres. VK7BO leaves his receiver running during the night. It uploads spots to the internet so I have direct feedback. As long as the band looks like being open, I transmit periodically to gather SNR values. Results confirm earlier findings of enhanced propagation during twilight.

Meteor scatter
One of the first applications that WSJT was intended for, is meteor scatter. There were times where high speed CW messages were exchanged and tape recorders with different speeds were used to slowly play bursts, to decode the CW message. With JT, this has become a lot easier, as the software does the job of coding and decoding very efficient. One of the latest developments is MSK144, which I like very much. Its predecessor, FSK144, had the disadvantage that the decoded stream contains a lot of errors and the operator has to determine if the messages make sense. With MSK144, the coding scheme contains error correction bits and the result is more or less binary: decoded or not. This again creates new possibilities, like automatic reporting and auto sequencing.

I have been using MSK144 for some time now and I can only say that it works tremendously well and resulted in a number of interesting contacts on 6 and 4 metres. The auto sequencing is also very efficient, the computer advances to the next step before you know it, so no time is wasted with repeating a message that was already received at the other end.

WSJT, WSJT-X and forks
Meanwhile, different versions are available. The software is open source, so others can create their own versions. Popular forks are JT65-HF and JT65-HB9HQX. The latter has many nice features. Currently, I mostly use JTDX for 60 m and WSJT-X (1.7 alpha version) for MSK144. LZ2HV has created a very good application for Meteor scatter, that is much appreciated by meteor scatter fans. Because of the rig control features in WSJT-X, I have been using that so far.

JTDX has additional decoding options, with which I have seen decodes down to -29. With  WSJT-X, the practical limit is around -22. Apart from the decoding margin, JTDX claims to perform better in crowded situations.

Update 2 January 2017
During the holidays, I found time to conduct some tests. I can confirm that JTDX beats WSJT-X with JT65. Here is a report: JTDX compared

Cheating?
There are those who claim that JT modes are close to cheating. Some even say it is.

To some extent, that may be true. Decoding is possible with help of a list of known callsign/locator combinations. So instead of reading a random sequence of characters, we compare with an already known reference. It is easier to recognise something we expect than to decode unknown information.

But if one claims that recognition is cheating, then we are doing that all our lives. As humans, we compare signals from our sensors (eyes, ears etc.) with known references. Languages are coding schemes. Characters as you read them here, is a way of coding. You can only decode, when you know how. So, you need prior knowledge to be able to decode. I consider this prior knowledge to be the shared key. Because a language is a matter of common understanding, which is exactly the pre shared key.

Getting back to the works of Joe Taylor, the “trick” is the amount of prior knowledge that is applied. We as amateurs know operating practice. We use Q codes, we use RS(T) reports, and so on, we simply expect certain information at a given point. What is the difference between a QSO with a DX station, of which you already have the callsign form a DX cluster, and a JT decode, based on a list of known calls? One might argue that the DX station does not know your call and has to decode that fully. But even that is not necessarily the case. There are many well known DXers (notice the well known part) and the DX station could recognise a call because it is known.

So at the end of the day, if those who claim that JT decodes are cheating, they may have overlooked that traditional contacts in the pre-internet era, follow exactly the same principles as JT modes, by recognising something we already know, albeit to a lesser extent. It has to be noted, that a transmission of something we already know, contains no information!

It is up to everyone self to decide where the dividing line is. For me, the principle that a two way contact over our airwaves has been established, should be leading.

For those who regard that as insufficient, they have the freedom of choice to select other modes and – even better – switch off all (computer) assistance.

For me, Joe Taylor and the developers around him have opened a new range of possibilities, including studying the ionosphere and which provides me a lot of pleasure just as well. Seeing your modest signal from the back yard detected at the other end of the globe on 60 metres, is something we very likely would not have known without Joe Taylor!

Thanks so much, Joe, for all your effort!