My 80 m version is shown below, but right now it is very popular with some incredibly cheap Chinese ones on sale on Ebay and other places.
The Pixie 2 uses the versatile LM386 amplifier for its audio output. I have shown previously on this blog how its gain can be boosted and how it can implement a CW filter, and also how the muting can be improved. However, during transmission, the LM386 just sits there idle, although it can be used to amplify a sidetone from an external oscillator.
But I'm sure the old 70's LM386 can do better than that. Despite its age, recently some pretty amazing uses of this chip have been demonstrated. It can be used as a regenerative receiver at least up to medium wave frequencies and it can also be used as an envelope detector/demodulator.
The LM386 challenge is this: Is is possible to implement a sidetone oscillator for the Pixie using only the LM386 with as few other components as possible? The output level needs to be controllable in order to make it comparable to that of the Pixie in the receiver mode.
The best data sheet for the LM386 seems to be the one for NJM386 from New Japan Radio Co. It is, as far as I know, the only one which shows the various muting circuits including the one using pin 7 which I have explored. It also shows the LM386 as an oscillator: both a sinusoidal and a square wave one.
In order for the LM386 to be useful as a sidetone oscillator, I believe that the oscillation must take place in the input circuitry. That seems to be the only way to ensure that the output doesn't come out at a blasting full rail-to-rail swing as in the square wave oscillator example in the data sheet.
By the way, the data sheet referred to above is also the basis for the improved Spice model for the LM386 that just was developed. It came partly as a response to my complaint over how poor the present one was. Maybe the new Spice model, developed by EasyEDA, could help solve the LM386 challenge?