I think I see it. According to the schematic, R1 and C2 aren't
wired the way the Maxim data sheet shows. You have them connected to the
+5V bus, between the supply and the upper transistor. Maxim explicitly
requires a _dedicated_ track connecting these two parts _directly_ to the
drain of the FET. What's even worse, C3 and C4 aren't connected directly to
the drain.
Here's how I'd wire it:
Mount C3 and C4 on the ground plane with zero lead length, and merge
their hot-side pads with the FET's drain pad. If this is a multilayer
board, place the two caps so their hot ends merge with the FET's drain pad,
then drill _two_ ground vias for _each_ cap, on _opposite_ sides of the
ground pad. This form a shorted loop between the grounded pad and the
ground plane, reducing the series inductance between the cap and the ground
plane by an order of magnitude. Now bring the +5V supply to the hot-end cap
pads, on the _opposite_ side of each pad from the FET. This forces all
supply current to go by way of the caps before it gets to the FET,
minimizing common-mode inductance, and hence minimizing the switching
voltage fed back into the +5V supply.
Now run a dedicated voltage sense track from the FET drain pad to R1
and R2. This way you avoid the effects of IR drop in the +5V bus. If the
caps and the drain have merged pads, you could tap the pad on the capacitor
side, which should reduce switching ripple a little.
Another thing to watch is how pin 7 is routed to the FETs. I'd try
a dedicated run to the source pin of the upper FET.
Finally, you have to know the actual value of the upper FET's Rds at
operating conditions, in order to calculate the required value for R1.
That's one of the weird things about this circuit. What you could consider
doing instead, is put a low-value current sense resistor in series with the
drain, and then calculate R1 according to that resistor instead. This might
give better unit-to-unit repeatability.
Of course, I could be completely off the mark. I haven't studied
the heck out of the data sheet, and haven't done any bench testing.
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
