Thank guys for your replies .Some of them i have to study first

Let me make some things clear about the circuit and values chosen

1. I've measured turn-on , turn-off delay at 3.3 ms for the driver relay.All
results are
after calculating this delay .So what is see on the scope is at the moment
i'm explaining
2. the arc is across the driver relay .The power board is inside a control
unit box so
i have to leave with small distances and cpu disturbances.Its actually a
microcontroller
having the problem .Driver relay contact current rating is 5A at 220V
.Power relay coil is rated is 6
watts consumption at 220V.
3. Using large value capacitors over 33 nf was causing sometimes latch of
the power relay so i have
value limitation here
4. The capacitors i've chosen are X2 self healing 275VAC. With no ZC control
they are blown
to 0 nf value (some of them) after 10-12 months of operation.
5. I don't have the space or budget to use large sized capacitors rated at
higher voltages or SSR .
The idea was to use ZC to avoid using expensive and large size protection
snubber
So the question is .

Does the relay On/OFF time differs with time .If it's 10% it's not a
problem since again
the arc will not be so high .Since its the current break that causes the
arc i must switch off at Peak of the
ac voltage .This is what my reading confirmed .In this case switching a
resistive load must i change the driver
algorithm ???

Any helpful comments will be apreciated


See section 15.2.12.5 and in particular Circuit B of figure 15.6 in the
link below. You will want 450WVDC electrolytics and 600PIV rated diodes
along with possibly a small inrush limiting resistor in series with the
whole lot. The idea is to dynamically form a peak clamp circuit and this
has been used in many high end line surge protection devices for many
years with good success. The zero crossing helps a bit but is largely a
waste of time.
http://www.leachintl2.com/english/en...ies/00053.html