[time-nuts] IC for 1 PPS Output

SAIDJACK at aol.com SAIDJACK at aol.com
Fri Apr 7 14:00:03 EDT 2006


Paralleling CMOS inverters to drive time critical signals is not a good  idea 
because the inverters have skews between them, even if they are on the same  
chip (die).
 
The skew for an improved version of the 74AC04 part (the 74LVC04) is  
specified as up to 1.5ns! 
 
Interesting enough, this parameter is wrong in the Philips datasheet, they  
say 1.5ps (impossible, since the difference in lead size of a couple of mm 
would  already cause >>1.5ps skew). I confirmed with Philips that it's supposed  
to be nanoseconds, not picoseconds:
 
_http://www.semiconductors.philips.com/acrobat_download/datasheets/74LVC04A_6.
pdf_ 
(http://www.semiconductors.philips.com/acrobat_download/datasheets/74LVC04A_6.pdf) 
 
This 1.5ns skew would cause all sorts of issues on the edges since the  
rise/fall time I measure on these parts is <1ns, so theoretically you  could have 
one output drive high, and the adjacent one drive low for up to 0.5ns  causing 
a short circuit over the termination resistors.
 
It's better to use a single, faster, 5V compatible, LVC logic driving  
through a single series termination resistor. This series termination resistor  is 
adjusted so that an open-ended cable shows a perfect square wave on its end.  
The cable has to be about 2-3 feet long going into a >1MOhm or higher fast  
oscilloscope or FET probe input.
 
BTW: the value of the series terminator doesen't matter when the cable is  
50Ohm end terminated since there will be no reflection from the end of the  
cable. Thus the value of the series termination resistor will determine the  
peak-to-peak voltage into a 50Ohm end terminated cable. The gate needs to be  
choosen so that it can drive down to 50 Ohms without damage of course.
 
There is one advantage to using multiple gates: they will share the energy  
of ESD pulses, or external DC voltages applied to the connector, and thus have 
a  higher survivability under stress conditions.
 
bye,
SJ 
 
 



More information about the time-nuts mailing list