S?bastien Bourdeauducq wrote:
> What about putting to good use those ferrite beads which poisoned
> our lives earlier?

Heh, my wicked mind also wandered in the direction of an LC filter
to keep those greedy USB devices at bay :-)

But then I did the math and realized that, even with 20% tolerances
on all the caps, the voltage drop should only be around 200 mV in
the case of atusb, which by the way only implements Atmel's design
recommendation (page 189 of [1]):

"In addition it is highly recommended to connect a 10?F capacitor to
 the VBUS line

[1] http://www.atmel.com/dyn/resources/prod_documents/doc7799.pdf

The USB standard does allow a capacitative load of up to 10 uF.

The calculation:

M1 with 220 uF - 20%, 4.78 V on the "5 V" rail:
4.78*4.78*(220*.8)/2 = 2010 uJ stored in C178.

Now we add the 15 uF + 20% worse-than-worst-case (*) of atusb:
sqrt(2010*2/(220*.8+15*1.2)) = 4.55 V

That's just the caps sorting out things among themselves, without
even involving the power supply. So I'm not quite sure why it drops
by a whole 500 mV. Ohmic losses should be negligible on this time
scale.

(*) Of the 15 uF, 2 uF should only get connected when the RF side
    starts operating.

> We can also increase the value of the decoupling capacitor close to
> the USB port (currently 220uF for the two ports).

Yes, that can't hurt. I checked the BOM and C178 has a nominal
tolerance of +/- 20%, so this is good. (There are caps with quite
evil tolerances out there, so you can't always trust the nominal
value.)

I asked Joerg on #qi-hardware what he thinks about adding an
inductance and he said we'd need a true monster to make any
difference in this case. (I'm too lazy to do the math ;-)

What he mentioned we might consider is an NTC thermistor, though.
However, it's not entirely clear whether these would work well in
the scenario we have. They seem to be common in circuits
operating at mains voltage.

- Werner
_______________________________________________
http://lists.milkymist.org/listinfo.cgi/devel-milkymist.org
IRC: #milkymist@Freenode

Reply via email to