question: multiple crystals or one?

I chose to put 16 ohm limiting resistors on the mosfet drains. That keeps the mosfets within spec for Vds max when they both conduct.

I updated

I'd rather not mess with the GPIO line because of how much trouble I went through to get the tri-state sensing working right.

That should work. If it is too long you will get a distinct "tristate" period during a transition.

Le Mon, 08 Jan 2007 11:20:30 -0500, DJ Delorie a écrit :

Why not implement the two solutions in the PCB ?

2/ A simple BJT/Crystal based Colpitts oscillator followed by a single schmitt trigger gate (Fairchild NC7SZ14) that distribute clocks signal to each xin MCU pin.

and make some tests ... i personnaly prefer the second one. EMI issues depends on how PCB signal are routed (traces placed near a Ovolt low impedance plan).

Habib

There are one-wire serial protocols that are much less sensitive to clock precision. FM or MFM data for example. They can effectively do that calibration loop for every bit... probably unnecessary.

Tim.

Say what? After the clock goes through the baud rate divider there isn't anything near to that much variation.

Usually because it's auto-bauding, not because any particular baud rate is off.

There is if the *clock* has that much variation.

Not according to the chip's specs.

In perfect conditions async serial w3ith 10 bit symbols can handle 2.5% frequency error at each end (or 5% at one end)

Bye. Jasen