(My mail is failing, so this will arrive very late, but i'm sending it anyway
 because, well, i've already written it  ;-) It's nice to see other people have
 already covered the important points.

Congratulations! (It looks like you used every single pin on the
microcontroller ;)

(2008.12.24) kirk...@pdx.edu:
>> About 2.2 nC at our operating point.
> ...
>> I don't think it's necessary.  But after talking to Keith, we'll add
>> the footprints.  We've got the space and I've got to rework that part
>> of the board to fix the D/S swap anyway.
> Yeah, that's a really low gate charge. I couldn't find the current limit 
> for the CC1111 GPIO pins, but you're probably right in that a resistor 
> may not be needed.

Assume the current limit on CC1111 is tens of milliamps. The horror and
hash on the igniter side is order amps into a ~1 Ohm load. I'm guessing
transition on the igniter side is no faster than ~100ns. Reverse
capacitance on FDN335 ~40pF, so capacitive feedback impedance >~2500 Ohm.

Reflected hash should be under 10mA, probably ok.

During check-out 'scope the gates during ignition. If there's
oscillation or hash then a resistor in the 10-100 Ohm range should fix
it. IMO don't go above ~2k, because at that point the switching
transition will start to slow down.

>> I hate to take the voltage drop hit on a 3.7 volt nominal battery system...
> Agreed. In this type of application I would opt for a P-FET such as the 
> FDS6375, which keeps the voltage drop to the low millivolts or less. In 
> this case, you *would* want to swap the D and S connections to that the 
> body diode conducts only when the battery is right, and even then only 
> until the P-FET powers up. However if you can get it done with just the 
> connector, great! In my applications (off highway industrial equipment), 
> reverse battery connection is common, so I'm always thinking about it.

Some Li+P batteries these days have reverse protection circuits in them.

If the battery polarity was successfully reversed, i'm as confident as i
can get without trying it, that stuff will smoke.


I'm not seeing anything that really requires change. What follows can
probably be safely ignored.

I don't put traces as close to the board edge as you have shown. Same
for component U4. Typically there would be a design rule of perhaps
1.5mm between components/traces and the board outline. This can be
handled at saw-out rather than in the gerbers.

I'm not sure to what extent your RF layout follows TI, but for example,
the hole between silkscreen indicators C23 and C24, what will surface
tension do to those adjacent components? I'm not sure, but it doesn't
look to me like they'll come out quite straight. But maybe i'm wrong.

Typically manufacturers don't recommend full solder paste coverage of
the exposed pad on parts bigger than ~3-4mm. They want to give the
solder paste farts somewhere to dissipate. The CC1111 data sheet
(SWRS033G p41) has a similar recommendation. I haven't seen the
tic-tac-toe pattern shown in the datasheet before. Seems like it would
work. All i typically do is reduce the paste coverage to ~60% and leave
the mask alone, but i don't have any high-volume experience, maybe their
way is more reliable?

Finally, for the record, i will lobby for a change that i don't think
you should or will make.

I don't like mechanical power switches for high reliability devices, nor
for rechargeable batteries.

If it were me, i would change the power jumper to a 3 position. Jumping
1-2 would set a flip-flop or something, and the device would be on.

The device could power-down under software control at any time
thereafter, but only two circumstances would normally apply:
  1) Battery voltage falls to a minimum threshold
  2) Jumper moves to position 2-3 and remains for more than 10 seconds
     AND the state machine is not in any flight state.

I haven't designed it but i think this change would cost an extra jumper
pin, a single gate flip-flop, probably one small FET, and a couple
resistors/capacitors. Since you're out of microcontroller pins, some
creativity might be required, but there's most often a way.

psas-avionics mailing list

Reply via email to