Dan Michaels email (remove spam text)
Ryan Miller wrote:
>> 1. EMI and spike filters at the hiVAC power input
>> - series EMI filter, MOVs, etc.
>Could you give examples of the EMI filters. I assume you're
>talking about something that goes across the contacts of the
>relay switching the AC.
Actually, I was referring here to general filtering on the
power input from the AC line - which you may already have - and
which any industrial controller should have. Many commercial
versions are available, usually a C-L-L-C filter across the lines,
[C's across, L's in series in each line] with 2 C's to earth gnd,
plus MOV on downstream end, etc. Among other things, it will
help protect the controller from external noise [lightning, etc],
as well as spikes generated by the switched loads and which are
conducted into the controller via the AC lines.
As mentioned, directly measuring the transients across the relay
contacts during on/off will help identify ineffective snubbers.
Alternatively, write some test code to exercise the relays
repetitively [at 1/sec or so] and observe which loads cause
problems. Fix those snubbers/etc. Probably use a smaller R.
By "signal" lines, I really meant "all" of the low-level lines
in and out of the uC [bad terminology on my part]. It is common
for spikes to be conducted from the switched circuits directly
backwards to uC pins driving the relays. Series R's and RC's
will help here. Also, I don;t recall transorb C values, but I
believe they are small enough to not have to worry about in the
comm lines - specifically regarding the *low* V transorbs.
Here, the best defense is a good offense. First, try to kill
the spikes as dead [ie, make small] as possible. After that,
it's a systems-level, "defense" problem - ie, add preventive
measures so that spike residues have as little effect as possible.
I try to err on the conservative side, and include as many
good design features as possible, rather than just try to
get by with the minimum allowable. I think the extra cost
and effort helps in the long run. The guys I mentioned
consulting for had 10,000-20,000 units in the field and a
significant amount of "warranty" costs - about 2-5% failures
per year - that's of a lot of $$$$. By working on the
protection on the controllers over time, we were able to
greatly reduce the in-field failure rate.
>> 7. Use watch dog timer in uC s.w. Use debounce routines on
>Watch dog on the pics (2 17C44) as well as cpu supervisors
Man, you already know/have it all.
Reset vs latchup is, I think, mainly related to the severity
of the garbage picked up. Your's sounds like it's on the lesser
side of real evil. The guard rings might have some small effect,
but will probably not produce magic. 1st attack the source of
the spikes - as above.
From my experiences with actually doing this, I think a good
patch of duct tape is about as effective as a real gnd plane of
the same area/etc. If adding a guard ring will have any significant
effect on the problem at all, I think it will be highly noticeable
with the duct tape jury rig.
Re ground used - try both ways & see what happens. Earth gnd might
be better. Also, it would probably help to keep the uC and relay
gnds separate if possible - if using an ULN2803 type driver or
BJTs/MOSFETs, you could run the gnds separately to the tie point.
- Dan Michaels
See also: www.piclist.com/techref/index.htm?key=120vac+traces
You must be a member of the
piclist mailing list
(not only a www.piclist.com member) to post to the