QRP Labs QCX 5W CW, Assembly Instructions Manual, Page: 85 / 138

QCX assembly Rev 1.08
85
4.23 Decoder menu
The Decoder menu contains a number of configuration parameters relating to the CW decoder,
which are described below. Some of these parameters control some aspects of the decoder
behaviour. Some constructors may find it interesting to experiment with these settings and see if
you can improve the performance of the CW decoder in your specific circumstances. For example,
some stations may experience more noise interference than others, depending on your location
etc.
5.1 Noise blnk.
10
This parameter defines the duration of the noise blanker in milliseconds. The microcontroller’s 10-
bit ADC samples audio at 12.019 samples per second. 48 samples are analysed by an
implementation of the Goertzel algorithm (kind of a single bucket of a Fourier Transform), which
results in a digital filter bandwidth of 250Hz. In other words, it results in a measurement of the
amplitude 250 times per second, i.e. once every 4 milliseconds. The amplitude is analysed by
logic which compares it to a threshold amplitude to decide if a tone has been detected or not.
Impulse noise that generates shorter pulses than the noise blanker parameter, is ignored.
If the noise blanking period is too short, then noise impulses will not be blanked effectively. On the
other hand, if the noise blanking period is too long, then it will impair the decoder’s ability to
decode high speed Morse. For example, 24wpm Morse has dits lasting 50 milliseconds.
5.2 Speed avg.
07
The duration of dits and dahs is measured in order to define a threshold at which to define a tone
burst as a dit or a dah, and whether to define no tone at all as an inter-symbol, inter-character or
inter-word gap. The measurement of this timing is implemented via an exponential moving
average, whose averaging duration is determined by this parameter (the weight of each new
measured symbol in the accumulated average).
If the exponential moving average is too fast (the parameter value is too low) then noise etc will
throw off the timing averages too easily. If the exponential moving average is too slow (the
parameter value is too high), then too many characters of the other station’s transmission will be
missed, while we try to adjust to the speed of his sending. This can be particularly offensive in
some contest or pileup situations where exchanges are very short.