[LinuxPPS] kernel 2.6.32 - ntpd-4.2.6 linuxpps experiences

[clemens at dwf.com]

> On Sat, Dec 19, 2009 at 18:29,  <clemens at dwf.com> wrote:
> > This behavior is not strange at all.
> > Each driver sets its own desired combination of ASSERT/CLEAR.
> 
> great, Reg, thanks for clarifying this.  It makes sense, though my
> limited knowledge of the matter does not allow me to understand the
> subtle difference.  Why would the Oncore driver would kill the CLEAR?
> It is a matter of IRQ load on the machine, or what else?
> 
Well the signal is 'kinda a square wave, actually its more like 20-80 on the 
ONCORE.

The assert signal is accurate to within nanoseconds of occuring on the second.
But now you have to get the signal back down to zero, for the next second.
In the ONCORE Hardware (and many others) this is just done with a RC timing
circuit, so its position is not accurate.  So the code is going to ignore it.
In some hardware the ASSERT is so narrow that the serial hardware in your 
computer
cant see it.  In these cases you will see reference to 'pulse streching' 
circuits, which
are again just RC circuits to make the pulse longer...but leave the leading 
edge as
accurate as possible.

The choice of turning of the CLEAR signal is one of simplicity.  Back a couple 
of
months ago the linuxPPS was returning both ASSERT and CLEAR no matter how you
had things set.  One simple 'if check' allowed this to work by ignoring the 
CLEAR
signals.  But turning things off is simpler.

Hope that helps.
-- 
                                        Reg.Clemens
                                        reg at dwf.com