Re: Making an X-10 lamp module immune to dimming

[isw <isw@xxxxxxxxxxx>]

In article <ldmdnSl8M_e_q9DVnZ2dnUVZ_g2dnZ2d@xxxxxxx>,
 "Robert Green" <ROBERT_GREEN1963@xxxxxxxxx> wrote:

> "isw" <isw@xxxxxxxxxxx> wrote in message
> news:isw-21C5D4.22184707062008@xxxxxxxxxxxxxxxxxxxxxxxxx
>
> > > <stuff snipped>
>
> > > So it's not possible to have a sine wave that has four "bands" of on/off
> > > pulses?  Or is it just that way for the X-10 modules?  I thought, when
> you
> > > talked about pulses, that each one of them turned on the triac for a
> tiny
> > > bit of time, and that all of them added together in each cycle increased
> the
> > > power output.  That would require switch the triac on and off multiple
> times
> > > within the cycle.  That description is true of how all the chopped A/C
> > > cycles add up, but I think you're saying the pulses that follow are not
> > > having any effect on the state of the triac.
> >
> > Correct; they do not. Once a triac is conducting, it's very difficult to
> > get it to turn off any way except removing the voltage across it
> > (actually, it's removing the current through it that makes it turn off).
>
> Are there other devices that can interrupt current flow with a gate trigger?

Yes. There are gate-turn-off (GTO) thyristors:
http://en.wikipedia.org/wiki/Gate_turn-off_thyristor

> > Right on. But think about how a half-sine looks -- a linear increase in
> > on-time does NOT result in a linear increase in power delivered to the
> > load.
>
> Right.  This looks very much like an "area under the curve" sort of problem.

Close, but more complex, because the load (assuming an incandescent
lamp) is not constant. The resistance of a lamp filament is lowest when
it's cold (which is why a burn-out so often occurs when it's switched
on), and rises as the filament heats up.

> > > So it might even be possible it's really just noise with no particular
> > > function but no particular harm to the operation of the module.
> >
> > It's not noise; it's a regular string of pulses.
>
> I thought that any signal other than the one you wanted could be considered
> noise.

That's a simplistic definition, although it sometimes works.


> > I didn't care about repeaters. What I wanted to do was control a large
> > number of electrical valves for an irrigation system. As I said, the
> > problem was that it was not repeatable -- the same number of "brighten"
> > pulses did not always result in the same pulse width, and so I could not
> > reliably select which irrigation circuit I wanted to run.
>
> Too bad.  What's a large number?  There's a lot of headroom with 256
> discrete X-10 addresses, even if you're controlling a modern home.

That National chip (an LED bar generator) has ten outputs, and it can be
hooked up sort of like a ten-value A-to-D converter. Having to use an
X-10 module for each and every sprinkler valve would work, but has even
more problems -- not least of which is the tendency of the things to
turn on for no evident reason ("noise" if yo will).

> <stuff snipped>
>
> > > So with the capacitor, the signal is being smoothed out from a pulse
> train
> > > to voltage level.  Usenet II is going to need a whiteboard.  At least
> ten
> > > times now I've wanted to show something with a simple sketch.   What
> does
> > > that pulse train look like?  Is it a square wave with the lowermost part
> at
> > > 0 volts and the uppermost part at Vcc+ (not sure if that the right
> term).
> > > Would the resulting DC level be Vcc+/2 or something close to that?
> >
> > Honestly, I don't remember. I was doing this some years back. All I
> > recall is that starting from "full dim" there was one narrow, "square"
> > pulse, very close to the following zero-crossing. As "brighten" pulses
> > were sent, that pulse always moved earlier in the half-cycle, but as it
> > moved, it was joined by progressively more and more that looked just
> > like it. The "extra" pulses occurred later in the phase than the one
> > that turned the triac on, and so they had no effect on the operation of
> > the unit.
>
> Interesting.  I assume that with X-10's parsimonious Scotch heritage that
> they decided that as long as it didn't have a negative impact on
> performance, they didn't need to add any circuitry to filter it out.

Yup.

> > Ripple would be a good term to use. The trade-off is that as you
> > increase the time constant to reduce the ripple, you also increase the
> > time the unit takes to respond to changes in the pulse width.
>
> This sounds like one reason that circuit designers are always tweaking the
> values of capacitors and resistors between board versions:  to optimize the
> various design trade-offs.  (-:

Well, yes, but usually a *real* designer will have a good notion of what
values to use and not have to try a bunch of things to see what works.
Changing a board layout can slightly alter things like the capacitance
between adjacent traces, and that can cause instability, especially if
the original values were determined "experimentally" by someone who
didn't understand worst-case design, margining, and so on.


>
> > > > > > 4) Change the 330 ohm resistor to 2.2 k
> > > >
> > > > Simplistically, the 12K resistor determines how fast the cap charges
> > > > (but not discharges, because the diode prevents that), while the 2.2K
> > > > controls the discharge time.
> > >
> > > Is that capacitor discharging during the time the pulse is at zero, thus
> > > providing current in the circuit where there would have been none
> without
> > > it?
> >
> > Yup.
>
> It seems as if the R/C circuit is one of the most common electronic
> "building blocks."

Yes, it is. Both components are cheap, and easy to get in a wide range
of values.

Here's an interesting item: Of all physical units (weight, volume,
current, voltage, heat, ...) resistance is available over by far the
widest range, easily from under a micro-ohm to over a hundred megaohms
-- say fourteen orders of magnitude. That's why electrical simulations
of other phenomena are so easy.

Isaac