I ran more simulations on the model I have from work with John Z.? I also adjusted the BV (breakdown) value of the BS170 model to be 1,000 volts instead of 60.? The model results were supported by bench tests of a Rev 1 QDX built for 12 volts.
1 - The voltage spike on the model is 202 volts.? This is confirmed by bench testing that the spike can be over 100 volts.? Variations in the actual breakdown voltage of a real transistor would give different results with different BS170s.
2 - The commutating diode and the Zener option reduce the spike to Zener levels or less with the commutating diode.? This was also confirmed by bench testing
3 - The power delivered to the load was reduced by about 1/4 watt for only the commutating diode.? This was validated with bench testing to be higher at approximately 1/2 watt.? No change in power was measured with the Zeners.
4 - I ran harmonic scans on my QDX for both solutions and found no difference from an unmodified QDX.
I did not do efficiency models like Hans, but I agree that the commutating diode solution impacts efficiency and power out.? Both seem to protect the BS170s from excessive voltage.
I have a theory that L14 is acting as a storage device that pumps the current each cycle until equilibrium is reached.? This is shown on the model as a gradual buildup of the voltage and power output and on the drains of the BS170s.? This is supported by an article in which John Z pointed me to current fed push-pull DC-DC converters.? I lost the link, but you can look up similar articles online.
Based on the above, I am using the Zeners to limit the potential voltage damage of the BS170s.
73
Evan
AC9TU
