Critique my PCB/Schematic!

[perky]

For the sake of a P-channel FET and a pull-up resistor I think it would be a good insurance policy ;-)

[TomWS]

For the sake of a P-channel FET and a pull-up resistor I think it would be a good insurance policy ;-)

Since the voltage is greater than the processor's, you'll need an N-channel FET to drive the gate of the P-channel FET.  Given that our Mr snorp wants to use through hole components, there aren't many convenient choices for these components. 

Now if we could convince Felix to take his SOT-23/SC70 breakout boards, keep them panelized, and populate a bunch of them with a load switch (eg Si1869DH) then our Mr snorp would have a viable through hole option and everyone would have a useful load switch for their breadboards...

I'm not convinced the leakage will be an issue in practice.  I'm not questioning your advice, it is valid given the datasheet, however, I suspect the leakage is measured with no connection on the inputs.  If these are solidly pulled to ground from the 328P outputs, even while sleeping, then I think it unlikely there will be much leakage and it is also unlikely that this circuit will be called upon to operate at 85  degrees C...

Tom

[perky]

Whoops, I forgot the Moteino had a 3.3V regulator! Yes, you'll need a bottom N-channel FET as well. I prefer the FDC6420C for load switching being in a SOT23-6 package.

You're right, normally it would be the Iceo spec with an open base, this means leakage current flowing from collector to base has no choice but to flow through the base to the emitter, and that is current gained to form the total collector current.  They've mixed up the terms a bit on that spec I quoted, Icex is usually the leakage through the collector (actually technically the emitter) with a reverse bias across the base-emitter junction, but they have appeared to quote Iceo as shown in Figure 1A.

Anyway, there's also the possibility of leakage through the base (the Iin(off)) which might also cause issues then the input is grounded. It's hard to figure out exactly what the leakage currents are going to be. One thing is for certain, this device isn't really designed for micro-amp current applications. I'm wondering whether discrete N-channel FETs would be better, it'll be more efficient and you won't lose the 200mV saturation voltage, and input impedances will be very high. Just a thought.
Mark.

[snorp]

If I get a "bad" chip that for some reason has more leakage than what I've seen so far, is that something I could just test when I build the board, or could it start to leak more over a period of time?

I noticed in the ULN2003B they added some pull-downs to avoid leakage. Maybe that would be a better choice? The ULN2003LV also has pull-downs on the input to allow tri-state usage, but it looks like it won't handle the power I need.

[perky]

The B version looks better, but I'm still dubious. If you're using a high-side switch the issue goes away ;-)

BTW if you ever do go for discrete FET approach you'll also need some diode protection, here's an app note that's got some useful stepper motor stuff including a FET based drive circuit:
https://www.silabs.com/Support%20Documents/TechnicalDocs/an155.pdf

Mark.

[TomWS]

I prefer the FDC6420C for load switching being in a SOT23-6 package.

One advantage of the FDC6420C is that the transistors are not 'pre-connected' so you can insert a resistor between the N-channel drain and P-Channel gate to control rise and fall times independently.

Of course, the SOT23-6 is HUGE, but if you're hand soldering... 

In either case, we should talk Felix into making these...

Tom

[perky]

Yes, you can add a resistor in the P channel gate drive to slow it down if needed (e.g. for a large capacitive load), also along with that adding a small cap between the switched output (i.e. P-channel's drain) and the P-channel's gate works nicely. Nice little app note here for this:
www.onsemi.com/pub/Collateral/AND9093-D.PDF
Mark.