While waiting for the first prototype board for the FE-5680A discipline board, I've been playing with the design for the follow-on.

One thing that bugs me a little is using two different buck converters - the LM3485 for the +15V supply and the MC34063 for the +5V supply. I picked proven designs for both. The LM3485 design, in fact, has successfully powered an FE-5680A. My old breakout board used an LDO for +5, but it was at the hairy edge of acceptable power dissipation (because of the 10 volt drop) with just the FE-5680A. Adding all of the GPS machinery would be way too much. But it's not so much current that an external switch is necessary. Thus the choice of the MC34063, which has its own built-in switching element.

That switching element can be used to switch an external element when the current exceeds the internal switch's capacity. That's necessary for the +15V supply because the peak current is 4A - far too high. Switching an external MOSFET is not completely trivial, however. Switching it on and off needs to balance the switching current against the switching speed. Switch too fast and the switching current may exceed the 34063's switching current spec. Switch too slowly and the MOSFET's internal power dissipation will be too high (the longer it spends in the middle ground between on and off, the more power it dissipates). Another issue is that the MOSFET's Vgs is ±12 volts. So the design is going to have a voltage divider between the MOSFET's source pin, it's gate pin and the 34063's switching collector pin. The switching emitter pin will be grounded. The intent of the voltage divider is threefold: 1. when the 34063 switch is off, the gate will be dragged upwards to insure the MOSFET is switched off; 2. the Vgs limit will be observed; and 3. the switching speed will be slowed slightly to balance switching speed and switching current.

I originally went down this road just because I wanted to simplify the project's BOM by sticking to one switching controller. But I like the MC34063 better than the LM3485, because the latter's hysteretic design means that as you raise the input voltage, the switching frequency rises. In the original Pi Power circuit, this led to problems running at high input voltages, leading me to reduce the maximum input voltage down to 14 VDC - lower than I would have otherwise wanted. The MC34063's switching frequency is fixed (and tunable with the timing cap).

Fortunately, I'll be able to experimentally switch out the resistors at build-time to tune the design if something starts heating up.