Close

Change of architecture

A project log for Yet Another (Discrete) Clock

I HAD to finally do this basic "exercice de style" in digital electronics, using some hundreds of transistors and diodes...

yann-guidon-ygdesYann Guidon / YGDES 04/17/2016 at 03:560 Comments

The crystal tube is a success !

https://hackaday.io/project/10698-clockwork-germanium/log/36003-it-just-works

18KHz is not a standard clock frequency. 32768 is very widespread and might be more convenient but I'll stick to it for this project as well, for at least 4 reasons :

  1. It works well. Oscillations are not as difficult to get and maintain than the typical wristwatch tuningfork. There are much better chances to make it work, which is important for DIYers with very limited equipment. (Wait, maybe the tuning fork didn't work as well because of a mismatched feedback capacitor)
  2. Since it works better, it draws less current. It's energy saving, so it's better than the discrete tuningfork oscillators I've tried until now. It has a place in this project too!
  3. Since it's a lower frequency, the crystal is larger and there is less pressure for high-speed design.
  4. It looks cool. Very cool. And people expect it to glow ;-)

Now I need to adapt this circuit to the MOSFET technology.

I'm concerned about the temperature-induced drift so a complementary design seems required. This will not compensate for gain drifts but at least reduce the polarisation/working point drift, unlike an asymmetrical design.

The power supply will be regulated at 2.5V, which according to https://hackaday.io/project/9376-yet-another-discrete-clock/log/32133-power-supply-voltage will draw almost no current. As an exception to the design principles, I'll regulate this with another integrated circuit followed by a high-order low-frequency low-pass filter.

Will the following circuit work ? We'll see...

I don't see how to tune the drive strength though. Maybe I'll have to modulate the power supply voltage.

Concerning the startup: the oscillator will probably work at low amplitude, to reduce distorsion and increase accuracy. These conditions do not allow it to start up easily (or it may stall). A capacitively-coupled germanium rectifier might detect the oscillation, accumulate them in a small capacitor and control the power supply voltage...

Startup could be faster with another hysteresis-inducing resistor at the source of one of the sense transistors but hysteresis makes the system non linear and less accurate... For prototyping, a trimmer at the source will help manage the gain.

Discussions