Japanese translation of these modifications, by Shig JA1XRQ
Unreliable microcontroller start-up at power-on, solution by Peter G3JRH
Power Amplifier instability, solution by Daniel Ekman
LCD contrast fluctuations on key-down: solution, by Peter HB9TVK
LCD contrast adjustment improvement, by Peter DL6DSA
Moving the CW filter center frequency (with spreadsheet by Paul KE7HR)
Quite a number of QCX constructors have reported that the microcontroller does not start-up reliably on applying power to the kit. They have to switch off the power, and then quickly switch it on again shortly afterwards, to get the program to start running (and showing text on the LCD).
Peter G3JRH is a chip designer and has seen similar issues before. Peter stated that his QCX also did not power-up correctly, and he has investigated in detail. He says:
"It is not really a question of the slow rise on DVcc pin, but it is that the AVcc pin rises more quickly as it only has a 100nf capacitor to charge. This violates the requirement for the AVcc not to differ from the DVcc by more than +- 0.3V (see ADC converter section of the ATmega328 datasheet). I see 2 solutions to this, either place a 470uf capacitor on the AVcc pin or better still place a small schottky diode between the AVcc and DVcc pins (cathode to DVcc anode to AVcc), I have done this on mine and it now starts correctly every time. What happens when the 2 Vcc’s differ is that one or more of the internal diodes gets forward biased and the silicon latches up, even the reset pin does not work. I suspect that the reason some batches of 328’s work and some do not is simply minor processing differences."
I believe Peter's comments and suggestion make a lot of sense. In summary, fit a Schottky from processor pin 7 (cathode), to processor pin 20 (anode). The diode cathode is the end with the stripe on it, and should go to pin 7. It is possible to fit this diode on the underside of the board quite neatly.
An EVEN BETTER solution, which is another way to keep the voltages on AVcc (pin 20) and DVcc (pin 7) equal is: remove inductor L5, and solder this inductor between pins 20 and 7. This method has the advantage of not needing any additional components. But, the disadvantage that you need to desolder L5, if you have already assembled the kit. Although I have never seen a power-up issue on any of my QCX kits, I implemented these solutions on a 30m version of the QCX, for the sake of the photographs for this page. It was easy to desolder L5 without damaging the board or the inductor.
The photos show the diode solution (left) - note the polarity of the diode indicated by the white stripe, to pin 7; and the recommended inductor solution (right).
Some constructors, seemingly tending to be the higher frequency versions of the kit, have reported a PA instability that resulted in one or more of the following symptoms:
1) high current draw on transmit
2) low or unstable power output
3) dead MPS2907 (in first batch of 500 kits, which used MPS2907 for keying transistor Q6)
The solution to this problem was first suggested and verified by Daniel Ekman SA2KNG. The problem is caused by the Transmit/Receive switch transistor Q5 being switched on during transmit and allowing interaction of the PA and receive band-pass filter transformer T1. Connect a 10K resistor (exact value not critical) from the Q5 drain to +12V. This is shown in the circuit diagram fragment below. The easiest place to connect this resistor is under the PCB, from the Q5 side of capacitor C33, to the source of keying transistor Q6 which is at +12V. The section of the trace diagram shows the top right corner of the QCX PCB as it is normally orientated.
The following photographs show a 10K resistor installed. The resistor value is not critical. A spare 10K resistor is included in kit batches 4 and 5 (approx QCX serial number 1500-2500), for the purposes of making this modification.
This is not really a modification - just a solution to a problem which can arise. On Peter's QCX build the LCD contrast changed during CW keying on transmit. Peter's photograph shows what happens. One of the tabs of the LCD module is very close to (touching, even) one of the output Low Pass Filter toroids, L2. If you are unlucky, the tab of the LCD frame can cut into the toroid wire and make an electrical connection. Whilst the LCD frame is not actually connected to ground or anything else, you can imagine why coupling 5W of RF into the LCD frame could produce some unpleasant effects. It is easy to carefully bend that LCD module tab closer to the LCD module PCB so that it is clear of toroid L2. Then the problem is solved!
Several people have commented that setting the contrast of the QCX LCD is a rather fiddly operation. The reason for this is that the contrast potentiometer R47 has an adjustable voltage of between 0 and 5V at its wiper. But the contrast adjustment voltage of these LCD modules is always less than 1V. Therefore 80% or 90% of more, of the adjustment range of this trimmer potentiometer is "wasted"; the contrast adjustment is much too sensitive because it requires a tiny adjustment in the lowest 10% of the potentiometer's range. Peter writes:
"I have found a simple solution for improving the adjustability of the R47 contrast potentiometer in the QCX: By adding additional resistance between the grinder of the R47 and GND, the tuning curve of the linear potentiometer is changed so that the required low voltage (less than 1 volt) can be set much better. I used 6.8 kilo ohms, but the value is not critical. The advantage is that you do not have to replace R47 nor interrupt a track from PCB. The small change is done in a few minutes, if you attach the resistor on the solder side of the PCB, see photo."
I have produced a spreadsheet model of the modification, showing that the voltage range 0-1V is now spread across most of the adjustment range of the trimmer potentiometer so that adjustment will be much less sensitive. The curves are for 2.2K, 6.8K (as used by Peter) and 33K - showing that the resistance value really is not critical. Thanks Peter!
The CW filter in the QCX kit is based on the HyperMite design by David Cripe NM0S for 4-states QRP club, with some minor modifications. The CW filter has a centre frequency of 700Hz and a 200Hz bandwidth.
It should be noted that changing the centre frequency will also alter the bandwidth. So if you increase the center frequency for example, from 700Hz to 850Hz, the bandwidth will correspondingly be increased from 200Hz to 240Hz (approx).
Paul KE7HR wrote a spreadsheet for calculating required resistance value changes for a specified CW filter center frequency. You can download the Excel spreadsheet by clicking here.