transistor for high side load switch: turning off chips 3.3V ~100mA

[djacobow]

I am working on a design in which I'd like to switch off all power to a subset of chips on the board. Current consumption would be well under 100mA and this is a 3.3V design.

I'm looking at FDN306P ($0.21 qty 100), PMK50XP ($0.29 qty 100), etc.

Anybody have a favorite?

Switching characteristics are not important as this will be on or off for very long periods. I am concerned about on state resistance especially as the battery starts to fade and both Vds and Vgs start to go down.

[Richard Head]

Those devices you have suggested will work perfectly, but they are pricey compared to bipolars.
If it's not a cost sensitive design then by all means go for it. If it's a high volume product where every cent counts then I would be inclined to consider a PNP bipolar in a SOT-23 package.
The PNP bipolar should be a lot cheaper.

[T3sl4co1l]

If you're pressed for pennies, PNP might save you a few.  But you'll have to be miserly with the base current to keep battery life from being much worse.  Are those loads always at 100mA, or do they vary?  You generally size the base current for the worst case, and eat the excess as wasted efficiency at light load.

Oh, and don't forget to set all the interconnections going to those devices as 'Z' (tristate / input mode) or '0' (logic low) so you don't phantom-power them from the logic.

Tim

[Precipice]

When you say current is well under 100mA, is it far enough under to just drive from an IO pin? Two IO pins paralleled up and driven very carefully?
(If you're in a low power environment, take care about charging the decouplers on your switched side.)

[djacobow]

When you say current is well under 100mA, is it far enough under to just drive from an IO pin? Two IO pins paralleled up and driven very carefully?
(If you're in a low power environment, take care about charging the decouplers on your switched side.)

Nah, it's primarily a bluetooth radio that peaks at 75mA, normal xmit is 45-50mA and quiescently should be much less.

I will think about the BJT. Thanks for the advice, guys.

Will have to remember not to drive power through the inputs. I've seen it a zillion times. Chips can work all day through "magic."