MT63

MT63 is a digital for transmission in high-noise situations.

How it works

By encoding the data to transmit (what you type on the keyboard) in a complex way, using 64 different modulated tones, the MT63 developer Pawel Jalocha SP9VRC has been able to include a large amount of extra data in the transmission of each character, so that the receiving equipment can work out, without any doubt, which character was sent, even if 25% of the character is obliterated.

MT63 has the facility for a secondary channel running simultaneously alongside the main channel. This can be put to a variety of uses, such as the generation of a continuous identification or beacon.

The secondary channel is not a prime function of the mode and therefore some software provides for it and others do not. The option to transfer binary files, such as higher-level documents or spreadsheets, is similarly at the whim of the programmer.

MT63 is perhaps the most elaborate user of error correction techniques. It uses a Walsh function that spreads the data bits of each character across all 64 of the tones of the signal spectrum and simultaneously repeats the information over a period of 64 symbols within any one tone. This takes 6.4 seconds.

The combination results in superb impulse noise rejection. At the same time, in the frequency domain, significant portions of the signal can be masked by unwanted noise or other transmissions without any noticeable effect on successful reception.

Transmission speed is good because there are so many individual tones to describe the information, while the individual symbol rate per tone can remain slow (which is good protection against ionospheric disturbances).

Tuning of MT63 modes is not all that critical. This is because the mode can use Forward Error Correction techniques to examine different combinations of the 64 tones that calculate the correct location within the spectrum. As an example, MT63-1K will still work if the decoder is off tune by as much as 100Hz. MT63-2K is even less exacting, with an error of 250Hz being tolerated.

When to use this mode

When conditions are good and audio bandwidth is not an issue, MT63 (particularly the 2 kHz version) using the "long interleave" setting is the best mode to use. The data rate is very rapid and the multiple use of error correction techniques results in the most robust broadcast mode readily available to the amateur.

Although it can be a little tricky to identify by ear or eye, the mode has a generous tolerance to tuning inaccuracies and its immunity to impulse noise is second to none.

When signals are weak or unstable

MT63 becomes difficult to discern from the noise and the wider pass-band leaves the mode susceptible to admitting too many unwanted audio products. Under these conditions, MFSK16 hugely out-performs the other modes while still maintaining a respectable data rate.

While only occupying a modest amount of spectrum, the error correction and time interleaving combine to recover the data from the most marginal signals, a full 10dB lower in the noise than MT63 can manage.

One of the compromises with MFSK16 is the heavy burden of 100% power demand on the transmitter. It is also notoriously difficult to tune when barely visible on the waterfall display and very stable equipment is a prerequisite for successful operation.

When the channel is heavily congested with other traffic

MT63 and MFSK16 both cope well up to a point. However, if it is simply a matter of finding a small enough gap in the activity, then the minimal bandwidth of BPSK31 and QPSK31 are to be preferred. QPSK31 does offer better error correction but suffers with the need to detect four phase states rather than two.

This often means that little improvement is noticed in practice. The PSK modes are very easy to tune in and work reliably on modest computers.

Another great advantage of BPSK31 in particular is that it is the most well known and easily understood of the data modes.

For the future

A 3kHz version of MT63 would be very interesting to assess. This would be a good bandwidth to choose for optimum use of the space available under current licensing which continues to think of a single speech channel allocation as being this wide. In the broadcast application considered here, there would not be an issue if the arrival of the data were to be delayed by a few more seconds. Therefore, to improve accuracy without increasing bandwidth, it would be beneficial to add more time-based interleaving.

Not much exploration has yet taken place into some of the newer or more obscure variants of these data modes. For example, PSK125 is similar to PSK31 but by using two sidebands of 125Hz the data rate is dramatically increased. The greater bandwidth also allows the introduction of improved error correction. There is also PSK250, operating at a faster rate again.

However, the error correction employed does not approach that of MT63 and, as with all PSK modes, success is limited by the need for stable propagation conditions and, as the data rate climbs, increasingly powerful computing facilities.

Error correction

This technique is called Forward Error Correction. Other modes use FEC (for example AMTOR mode B uses a simple FEC technique), but MT63 has other advantages.

Unlike most HF modes where a character can be lost or changed into something else, by a single noise burst, MT63 is inherently very robust, because each character is spread over many tones (to avoid interference such as other radio transmissions) and over several seconds (to avoid bursts of noise, such as lightning).

"MT63's COFDM like properties"
* The MT63 signal is spread both in the time domain (temporally) and the frequency domain (spectrally). To ensure that noise bursts and other time domain interference artifacts have minimal effect, each encoded character is spread over 32 sequential symbols (3.2 sec).
* To ensure that frequency domain effects, such as selective fading and carrier interference have minimal effect, the character is also spread spectrally by using all the tones across the width of the transmission.
* On each of the 64 tones, the transmission data rate is fairly slow, which suits the nature of ionospheric disturbances.

Despite the low data rate, good text speed is maintained, because the text is sent on many tones at once. The system runs at several different speeds, which can be chosen to suit conditions, but 100 WPM is typical.

MT63 sounds unusual, (it sounds like a roaring noise) but the performance is spectacular. There is no connection process, as in AMTOR, Packet or PACTOR. Some users maintain that under poor propagation conditions (namly excessive fading) MT63 works better than either PACTOR II or Clover. Under good conditions the performance advantages are less obvious.

Jamming immunity

MT63 is also far more immune to interference and deliberate jamming than any of the more conventional modes.

In a "long interleave" option, the spreading is over 64 symbols (6.4 sec), with consequent improvement in resistance to impulse and periodic interference, but of course double the time taken for the data to "trickle through" the Walsh encoder and decoder pipeline.

Changeover from transmit to receive and vice-versa is however considerably slower than most modes. It therefore requires some skill and patience to "break in" on a conversation.

Bandwidth issues

There are disadvantages to MT63. First, the mode is broad (see below) and is quite aggressive, i.e. it causes interference to other modes, but itself is little affected by other modes.

Also, because of the delay through the error correction and interleaving processes, it is not possible have quick turnaround "slick" conversations. In other words, operation is clumsy.

Another problem with the reception of MT63 is the fact that correctly decoded text can take up to 15 seconds to appear on the screen. A clue to synchronized reception can sometimes be gleaned if the software has a digital squelch facility. If the squelch is set up to be closed when no MT63 signal is present, it can often be seen to open as soon as it finds MT63 and ideally the text will follow quite a number of seconds later! Another clue to satisfactory reception prior to the actual appearance of the text is that the generation of large volumes of unwanted characters tends to cease when receiver synchronization has been achieved. As with other multi-tone modes where many sine waves exists side by side, the transmission chain for MT63 must be made as linear as possible.

MT63 Latency

Latency is a measure of the time taken for transmitted data to pass through the transmission and reception equipment.
* It takes 165ms to transmit a 45 baud RTTY character via a UART.
* RTTY has symmetrical latetency so the total latency for it is 330ms.

Walsh coding: a design flaw?

The current 7-bit version of MT63 uses Walsh codes, with temperal interleaving. This may not be an optimal coding scheme with respect to error correction. It must be pointed out that turbo codes could be used to accomplish the same task, with a ~25% increase in efficiency.

Justification
* Walsh: 7 bits → 32 bits
* Turbo: 8 bits → 24 bits (the need for varicode can be eliminated)
* Walsh - Turbo = 8 bits, a 25% increase in system efficiency

The same temporal interleaving techniques could be used with turbo codes, but as a general principle only short length interleaving should be used with turbo codes.

Media

See also

* Walsh code
* varicode
* radioteletype
* shortwave
* PSK63

Related links

* http://www.qsl.net/zl1bpu/MT63/Technical.htm
* http://home.satx.rr.com/wdubose/HF/Mode_Comparisons.html
* http://www.qsl.net/zl1bpu/MT63/MT63.htm
* http://www.bartg.org.uk