Setup of an FM AFSK transmitter

This article explores optimal setup of an 1200 bps FSK transmitter.

The 1200 bps Bell 202 modulation scheme and transmitters

The 1200 bps FSK modems usually use Bell 202 tones at 2200 and 1200Hz.

Let us firstly characterise the ideal 'FM' land mobile transmitter. Though they are called FM transmitters, they have a pre-emphasis circuit that increases deviation at 6dB/octave over most of the speech band, certainly over the spectrum utilised by the 1200 bps FSK modems. Ideally, the deviation at the SPACE tone at 2200Hz is 5dB higher or 1.8 times that at the MARK tone of 1200Hz.

Another important system element is a peak limiter that appears in most implementations. It is usually between the pre-emphasis network and the modulator, so it has the effect of limiting peak deviation almost independently of modulating frequency, and at the same time is non-linear and so creates intermodulation distortion.

If sufficient drive is supplied from the modem, although the output of the pre-emphasis network may show the 5dB level difference mentioned above, the limiter (which is usually a peak clipper) will level the tones, ensuring approximately equal deviation, thus wiping out the pre-emphasis along with other intermodulation distortion that is like adding noise to the signal.

To recapitulate: AFSK overdrive wipes out pre-emphasis and adds intermodulation distortion noise to the transmitted signal.

Is that bad? Read on...

Receivers and demodulation

Receivers have a de-emphasis network to complement the pre-emphasis network in the transmitter. In fact, it is usually the case that the de-emphasis more that compensates the pre-emphasis, a cheat used to improve sensitivity specifications.

Historically, 1200 bps packet modems used mainly one of three chips:

They each use different techniques for demodulation, the AM910 using digital techniques, the XR use a PLL technique, and the TCM3105 uses a frequency to voltage converter.

Anecdotal evidence is that some packet modems suffer degraded performance when the audio equalisation is poor. That is understandable, the PLL approach is susceptible, and the frequency / voltage converter is especially susceptible.

Many problems attributed to modem techniques though are probably due to overdriven transmitters.

Is there a solution? Yes...

The solution

It is easy, don't overdrive AFSK FM transmitters.

There are several ways that one could go about the task of setting up the AFSK drive level on a transmitter, but firstly lets list some common ones that are not good:

These all fail, and the wide range of maladjusted signals on air is evidence of that.

To find effective solutions, lets consider the key components of the end to end channel, ie from modem to modem.

The end to end channel

The modem analogue tx signal is commonly fed into the microphone input of a land mobile FM radio. It will usually pass through a mic amplifier, preemphasis network, limiter, and frequency modulator. If such a radio has been adjusted for voice, the limiter is usually set to limit peak deviation to about 4kHz. The mic gain is usually adjusted so that clipping in the limiter occurs, but infrequently with normal speaking level.

The receiver has an IF filter that restricts the bandwidth of the signal fed to the FM detector. The output of the detector passes through a de-emphasis network and is amplified to drive a speaker. The speaker output is commonly fed to the modem analogue rx input.

It is important that the analogue channel from modem tx output to modem rx input is linear, delivers optimum S/N ratio, and has an approximately flat frequency response from about 500Hz to 2700Hz. The elements of the channel that are most likely to challenge that goal are:

If the tx limiter is driven to clipping, it distorts the waveform creating intermodulation distortion and degraded signal to noise, and it degrades proper pre-emphasis.

It the tx modulator is overdriven, it will create a spectral distribution with more of the tx power outside the receiver IF filter, which wastes transmitter power outside the channel and degrades S/N ratio.

So, the key constraints are:


There are many methods that can be applied effectively, depending on your knowledge, experience and equipment at hand.

A simple method that will work well in most cases

If a transmitter is already properly adjusted for voice, then ensuring that the AFSK drive level is as high as possible without any clipping in the limiter will usually achieve the desired result. The transmitter will operate at the maximum deviation without clipping, and that will usually be sufficiently low that most signal power is captured within the receiver IF filter.

Setup an independent monitoring receiver with an AC voltmeter monitoring the audio output. Do not try and do this setup by ear.

A lot of TNCs have firmware that supports a CAL command to transmit a steady tone for system line up. If you have one of these, use the CAL command (K to key up, spacebar to alternate tones, Q to quit, see your documentation) to transmit the SPACE tone, ie the higher frequency tone, and use that for the following procedure. (Hint: it times out so you have to reset it periodically.)

If you don't have a CAL facility, send a packet full of character 'U'. That creates a packet with almost equal number of MARK and SPACE elements.

Increase the drive level up from very low watching the voltmeter increase carefully for the first signs of gain compression, ie where the output does not increase exactly proportionately to the input. You are looking for just a hint of compression, then reduce the input just a little to be sure there is NO COMPRESSION, no clipping in the system.

Now reduce the drive for half the indicated voltage on the meter.

If in doubt, err on the side of lower level. Reducing drive level degrades received S/N ratio and decode rate, but not nearly as much as excessive drive.

Using a deviation meter

The first method focused on avoiding limiter clipping and assumes that the maximum deviation without clipping will be acceptable.

A better method is observe the onset of clipping using a deviation meter, and adjust the drive level with 2200Hz tone to the lesser of clipping onset and 3kHz deviation.



Version Date Description
1.01 06/01/2012 Initial.

© Copyright: Owen Duffy 1995, 2017. All rights reserved. Disclaimer.