Alexandre -

>    I have to control 8 triacs for dimming incandescent lamps with the
>minimum possible cost. I have been doing some research but there are
still
>some aspects I would like if someone could help me with. Follows the
list.

I've just finished doing my second project very similar to this.

First, I learned from the Teccor engineer that you should use a MINIMUM
of 6 amp triac to control lamps up to 1 amp (about 100 Watts at 120 volts
is actually what he said).  Bigger triacs for bigger lamps.  This is
because when a lamp fails, the filament can burn through in one place,
and the two pieces can then short out, placing a huge load on the line
(and your triac).  Fortunately, this doesn't last long because the
filament burns up again very quickly due to the excessive current, but
you need an oversized triac to survive this stress.  Or, you can use a
smaller triac and suffer the occasional smoke when a lamp fails this way.

>    - What is the best discrete zero crossing circuit ? I have found
many,
>ranging from a single resistor, a resistor and a zener up to a full wave
>bridge with clamping diodes and a transistor. Do they all work the same
in
>practice ? I guess they do not have to be very precise because I would
not
>be able to fire the triac exactly at zero crossing anyway ! Is the
>optocoupler circuit better ? I do not thing so, specially in my case
where I
>do not need isolation from the power line.

The best circuit depends on the application.  If you don't need
isolation, then don't use opto's.  They would be a waste of space and
money.  A simple resistor will suffice, but I usually prefer a 100K
resistor and 5.2V zener across the line, with a separate 10K resistor
from the zener to the PIC.  This provides better protection from
transients.  Either of these methods is actually pretty precise, and
would be fine for controlling lamps.  Noise on the line will cause some
jitter, but it takes a lot of jitter and thus a lot of noise to be
significant to lamp driving.  A 0.01 uF capacitor across the zener will
help reduce the noise, and add minimal delay of roughly 50-100 uS.  Yes,
this is much less than one time constant .

>    - How to fire the TRIAC without a Opto-diac ???? I have found many
>circuits and explanations again, but none has fully convinced me ! The
>circuits range from a single resistor from the microprocessor to a
push-pull
>stage with transistors and a capacitor to the triac gate. I will not be
>using a logic gate triac, so I know that I need more current to drive
the
>triac. Do I have to fire it with positive current in one semi-cycle and
>negative on the other in respect to MT1 ??

No, you don't need diacs.  They are used for timing - an R/C time
constant charges up to the diac voltage, then fires the triac.  You vary
the time by changing the R value (a potentiometer).  In your case, since
your timing is controlled by the PIC, all those circuits are unnecessary.


(All my references below are to Teccor specs.)

A resistor from the uP to the gate is fine for PIC's working with
sensitive triacs, and for the smallest non-sensitive triacs.  The PIC
can't provide the drive to turn a larger non-sensitive triac, since the
specs call for more than 25 mA.  You don't need push-pull, but you do
need some sort of high-current (>25 mA) drive.  You could use two PIC
pins with 2 separate resistors, but that uses up a lot of I/O pins.
You'd have to make sure you turn the pins on at the same time, or you get
partial gate drive for a while.  See below for why this is bad.

The reason you don't need push-pull, especially with non-sensitive
triacs, is that there is an intrinsic (part of the silicon) resistor from
gate to MT1.  Unless you have lots of noise and lots of lead length on
the gate drive wire to pick it up on, there's no worry about accidental
firing of the triac.

You don't need a positive pulse one time and negative another.  You can
fire a positive pulse all the time, or negative all the time.
Unfortunately, when the MT2 is negative, and gate is positive, this is
Quadrant IV.  Teccor doesn't spec how much current it takes to fire
non-sensitive triacs in QIV, but you can be sure it's more than any other
quadrant.  I'd stay away from unspecified regions if at all possible (see
below for how to do that).

>What happens if I am using 220v
>with 2 phases instead of a phase and a neutral connection ? My power
supply
>will be a capacitor coupled supply, without a transformer.

As someone else already said, using 220V is no problem.  Simply reference
your circuit to one phase, and it's no different than 120V with hot and
neutral, except that the voltage is higher (safety issues aside).

I would reference the HIGH side of my 5V to one phase, and MT1 to the
same phase.  That puts you in QII and QIII, where drive currents are
specified, and you can use a simple NPN transistor like 2N3904 to fire
the triac with negative gate pulses.  Or, if every penny counts, use a
2N3906 in common emitter configuration and eliminate the base resistor.

By the way, don't skimp on the gate drive if you want your device to be
reliable.  It may work great on the bench with 1/10 of the specified
drive, but you won't be turning on the triac quickly.  This stresses the
die, and reduces reliability.  You may also have problems at low
temperatures, where the drive requirement is significantly higher.

Since you are using capacitor-coupled power supply, getting enough drive
current for the triacs is a problem.  You want as short a pulse as
possible so it doesn't load down your power supply.  Calculate the
voltage required to push the triac's maximum hold current through
whatever lamps you are using.  Then see how long the line voltage takes
to get from zero to this voltage.  That is how long your pulse must be.
Don't forget to use worst case conditions - low-power lamp, low line
voltage, maximum hold current.  If you still can't afford a pulse this
long, then you can use a shorter pulse and delay it from the zero
crossing.  Or you could use several pulses.  The thing is that you have
to have a pulse that lasts at least until the time you calculated.  I use
two pulses, each 60 uS long.  One is as close as possible to zero
crossing, the other is about 1 mS later.

Contrary to what someone else said, you do not need 10V to trigger any
triac.  I've never seen one that requires more than 2.5V worst case.  You
just need to take this into account when calculating what resistor you
need to get the drive current you want.

>    Sorry to disturb with the OT question but I was not able to find
these
>answers anywhere else and I have already digged all the ap-notes from
>teccor, motorola and many web sites. It seems that everyone prefers to
use
>opto-couplers ! It does not make sense to me if I do not need the
isolation.
>The isolation will be in the connection betwenn the controller board and
the
>power board.

Horray for you!  You believed there was a better answer, and you stuck to
it until you found it.  That's what this list is all about.

You probably already know the rest, but...

One last caveat - when working on this system, be very careful, as it is
very dangerous.  An isolation transformer is highly recommended, and it
goes on your DUT (device under test), not on your scope.  DON'T connect
your scope ground unless you are using an isolation transformer - you'll
blow fuses, traces, thin wires, pieces of alligator clips, etc.  I've
seen it done (:-o)


Don