[net.micro] F245_oscillation

dean@felix.UUCP (06/12/84)

		FARCHILD 74F245 PRONE TO OSCILLATIONS




    FileNet Corporation has found that the Farchild 74F245 can cause
 excessive system noise under certain conditions.  Our system uses
 F245 transceivers to drive heavily loaded buses because the F245
 is fast and has good drive characteristics.  In two of our designs,
 F245s exhibited oscillatory behavoir not found in other TTL family's 245s.

   The condtions for oscillation are when the F245 is enabled driving
the "B" outputs (pins 11-18) and the "A" side (pins 2-9) are tri-stated.
The "B" side was terminated with 330/470 Ohm resistors on both sides of
the backplane.  Oscillations occur when the data into the F245 "A" side
are logic zero and the allowed to float from a tri-state condition.
As the "A" inputs float up due to the input current of the F245, the
"A" inputs will pass through the F245's logic threshold.  At this point
the F245 will break into oscillation due to an apparent feedback method
internal to the chip.  The "B" outputs will oscillate at about a 60 Mhz
sinusoidal wave at about 3 volts peak to peak.  Both VCC and ground 
have about a 1.5 volt peak to peak sinusoidal wave from this oscillation,
despite a .1uF bypass cap and excellent power distribution for the chip.


  The backplane for our system provides a low impedence path for signals
and the traces from the F245 to the board connector are very short. 
While this is good design practice, this seems to contribute to the
oscillation problem.  We did not notice these oscillations on our wire
wrap boards and a printed circuit board on an extender card does not
seem to exhibit these oscillations.


  The fix for us was to place pull up resistors on the "A" side of the 245
so that the inputs would not float near the threshold voltage during a
tri-state condition.


  We have found that the Motorola F245 behaves in a similar fashion to the
Fairchild part.  The oscillations occur for all date codes of Farchild
parts so this seems to be inherent in the design of an F245.


  You may think it is poor design practice to enable your tranceivers
with tri-stated inputs.  So you're right, get off my back, but you
should take a close look at your designs with F245s.