Olin Lathrop wrote:
>
> > A cap C-B in that circuit is stupid,
> > a cap B-E will fix his problem.
>
> Maybe.  It depends on where the feedback is coming from that causes the
> oscillations.  He said the relay was "vibrating" or something like that
> during the transition.  The circuit is just a DC amplifier driving the
> relay, and as you said, relays do have some inherent hysterisis.  We also
> know his input signal varies very slowly (outdoor ambient light at dusk).

Yes he said it was vibrating when the relay
was turning off, and that he had added a cap
C-B which Peter suggested. Which would cause that
problem exactly.


> So, why does it oscillate at the threshold input?  One possibility could
> be AC noise coupled back into the amplifier from the switching transients.
> Your cap from B to E should help with that.

The LDR and resistors are some 10's of kohms
and a 1uF or larger cap B-E would fix that due
to the large RC time constant, it simply can't
oscillate as the hysteresis of the relay is huge.
Only back-emf or as you mentioned supply voltage
rise could cause the oscillation on turnoff.
Both would be fixed by adding a cap B-E.
I also said if further filtering was needed to
add another cap on the lower resistors which is
basically a PI filter.

I've built many one transistor relay driving
circuits like this, they have enough gain for most
uses as the 12v relay pulls in at about 9v and
drops out at 4v or so. The low gain and slowly
changing analogue voltage won't matter.


> However, I suspect at least part of his problem is DC feedback via the
> power supply when voltages change due to the relay switching on
> significant power.  A B-E cap won't fix that, only decrease the
> oscillation frequency.

Yes I assumed that his PSU had a decent size cap
across it. Further decoupling the relay PSU from
the LDR PSU would help.

>  I think the best way to address both problems
> together is to spend a few more cents and add another transistor stage.
> This will give him more gain and provide an easy opportunity for some
> deliberate positive feedback in the amplifier to get some hysterisis.
> This will always drive the relay solidly on or off, and should provide the
> necessary noise immunity.  It will also make insensitive to small
> variations in light level at the thresholds, like a cloud momentarily
> passing over the sun shortly after it has set.

Sure another gain stage would be a step forward
but might not really be necessary. It also might
worsen the situation with clouds etc as the one
stage system will rely on large relay hysteresis
to mean that the relay won't pull in until a
large difference in light levels. I built a 1-tran
light relay once and had problems with too much
sensitivity, one transistor may be ideal for this
provided he has larger enough caps B-E.
-Roman

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.