QRP Labs QCX-mini, Assembly Instructions Manual

Page: 111 / 133

the sidetone were configurable in the software via the configuration menu. In order to
control the volume, the microcontroller adjusted the duty cycle from 50% (maximum
volume) down to under 1% (for minimum volume).
In firmware version > 1.02 and above, the sidetone generation method was changed. The
former method was a simple squarewave, with variable duty cycle, in order to change the
volume. However, this also changed the average level, leading to a DC bias at low volume
levels; on switching from Transmit to Receive (and indeed, vice versa), the DC bias through
the audio chain, being suddenly restored the nominal 2.5V, generated a large click.
The resolution to this was generation of sidetone by superimposing a PWM duty cycle
change, for example from 51% to 49%, on a much higher frequency squarewave (42kHz is
used at 700Hz for example; the higher frequency is given by 60 x sidetone frequency). This
duty cycle change at a rate of 700 times per second, creates a small 700Hz squarewave
superimposed on a full size 42kHz squarewave; the average duty cycle is 50% and
therefore no click occurs during RX TX switchover. The volume is changed for example,
made louder, by a larger change in duty cycle; for example 60/40% instead of 51/49%.
The sidetone is injected into the receiver path via a 3.3K resistor at the input to the CW
filter. The sidetone generated by the microcontroller is a squarewave, rich in harmonics. As
the volume is reduced, the duty cycle percentage drops and the amplitude of the 700Hz
fundamental tone also drops. There are many harmonics of course, and the CW filter does
a great job of removing these, so what is left in the earphones is a pleasant and pure 700Hz
sinewave. This is why the sidetone is injected at the CW filter INPUT.
During transmit, when the sidetone is operational, the mute switch Q7 is also closed – but
there is enough leakage through the imperfections of this switch that the sidetone gets
through anyway. The 3.3K sidetone feed resistor R59 is chosen to pump enough signal
through that it overcomes the attenuation of the mute switch. Without the mute switch, R59
would be a much higher value.
Key paddle, rotary encoder switch buttons
The microcontroller keeps an eye on all the buttons, key paddle inputs, and rotary encoder
switches. When button or switch closures occur as the operator activates a control, the
microprocessor responds immediately as required.
The paddle inputs, and the rotary encoder switch, are read using dedicated microcontroller
I/O signals.
All mechanical switches exhibit switch bounce, where the switch contacts generate multiple
transitions for a short time when the switch is activated. It is common to see in many
projects, resistor-capacitor networks to debounce switches (including the rotary switch).
Simple debounce circuits involving a resistor and a capacitor inevitably involve a
compromise when choosing the R-C time constant. It is easy to miscalculate and make the
time constant too short (bounce noise still gets through) or too long (rapid switch closures
are missed). In some cases it is impossible to find the sweet spot in between these two
extremes.
In my opinion, resistor/capacitor debouncing is a poor solution to the problem, when the
circuit contains a microcontroller. It is easy to debounce the switch edges in software! This
allows you to control time-constants or other
debounce logic much more precisely. Of
QCX-mini assembly Rev 1.05_Fr
111