Re: I think they've done it again.

[nobody@xxxxxxxxxxxx (Dave Houston)]

First, I screwed something up in translation. The consensus seems to be that
the total wattage connected to the switch needs to be about 300W or less and
the dim level needs to be about 50%. (Why higher wattage doesn't exhibit the
problem is a bit mysterious.)

Here's a very short Microchip App Note which discusses triacs, noise and
PICs.

     http://ww1.microchip.com/downloads/en/AppNotes/00958A.pdf

Here's another short app note from ST that discusses why IGBT dimmers do not
need chokes.

     http://www.st.com/stonline/products/literature/an/3735.pdf

Once switched on by a pulse on its gate, a triac can only be turned off when
the current through it drops below its holding current. This normally occurs
just before ZC but can be triggered at other points by noise un the line.
Here's another short ST App Note discussing this.

     http://www.st.com/stonline/products/literature/an/3563.pdf

Yes, a triac is a thyristor. But not all thyristors are triacs.

     http://en.wikipedia.org/wiki/Thyristor

Your plumbing analogy is not entirely inapt but things are reversed. Around
50%, the triac is being turned on at or near the point of maximum voltage.
This creates a "ringing" noise on the line (roughly analogous to your
pulsating shower head) very similar to the pulses that UPB generates by
discharging a capacitor onto the line.

     http://www.ieee-denver.org/comsoc/presentations/UPB.pdf  See pp8-10

PCS says the spectrum of their "data" pulse is in the 4-40kHz range. They
indicate thay got the idea for this when they noticed how well triac noise
pulses propagate through the "network".

Echelon says that the "ringing tail" of a triac noise pulse is in the 150kHz
range.


http://www.ipcf.org/doc/Echelon_-_Power_Line_Communication_Technology_Update.pdf

<SIDEBAR>
While researching the spectrum of triac switching noise I came across a
microcontroller project that, while it sheds no light on the current (pun
intended) topic, does illuminate (pun again intended) why there appear to be
~157 microdim increments.
     http://www.mikroelektronika.co.yu/english/magazine/articles/03.htm
</SIDEBAR>

"Robert Green" <ROBERT_GREEN1963@xxxxxxxxx> wrote:

>"Dave Houston" <nobody@xxxxxxxxxxxx> wrote in message
>> "Robert Green" <ROBERT_GREEN1963@xxxxxxxxx> wrote:
>>
>> >"Keep thinking Butch, that's what you're good at!"  :-)
>>
>> I'm not sure whether that's a reference to "Butch the Bulldog" or someone
>> else. ;-)
>
>Butch Cassidy and the Sundance Kid . . . I think it was right after the
>famous knife fight scene with Harvey (the 7 foot Lurch, from the Addams
>Family).  Butch uses creative thinking to win a potentially hopeless
>encounter:
>
>http://www.imdb.com/title/tt0064115/quotes
>
>Butch Cassidy: No, no, not yet. Not until me and Harvey get the rules
>straightened out.
>Harvey Logan: Rules? In a knife fight? No rules.
>[Butch immediately kicks Harvey in the groin]
>Butch Cassidy: Well, if there aint' going to be any rules, let's get the
>fight started. Someone count. 1,2,3 go.
>Sundance Kid: [quickly] 1,2,3, go.
>[Butch knocks Harvey out]
>
>(If you've seen it, you'll never forget it!)
>
>> A consensus seems to be forming among those seeing the
>> flickering that it only occurs when a switch has more than
>> ~250W (total) connected and is dimmed to 50% or less.
>
>So, this is a state where the triac is passing a relatively high amount of
>current per cycle.  Why would dimming below 50% be important?  Is that when
>the triac is likely to put "spikes" out on the line?   If electrical
>circuitry can be compared to water flow in pipes then quickly interrupting
>the flow of water, as in a shower massage, causes stress on all nearby
>joints and fittings.  Is the analogy even slightly on target as to how triac
>noise can affect other components ????  :-)
>
>> This puts the triac switching point near the peak of the line
>> voltage and at the furthest point (in time) from the Insteon
>> control signal. That the triac is being turned off at this point
>> is the least likely explanation as the current, being at or near
>> maximum, is almost certainly far above the triac's holding current.
>
>When the triac is triggered does it keep conducting until the current
>passing though it drops below the holding current or can it be turned off
>before the zero crossing is reached?
>
>(As an aside, IIRC, resistors change value when heated.  Do the properties
>of inductors change if they get very hot?  I'll bet those teensy little
>SwitchLincs get God-awful hot driving a large load with significant
>dimming.)
>
>> This is also the area where triac switching noise will be at a maximum.
>
>(Forgive me for breaking your message up like this, I know it's not your
>preferred style, but I need line-by-line guidance!)
>
>Why would the switching noise be at maximum?  Because the voltage is at its
>peak?  Because there's more area under the curve at the peak for a given
>length of time and so there's more current?  Is the noise created harmonic?
>
>> An undersized choke could allow this noise to reach the PIC
>> (in the subject switch or in other devices on the line) and this
>> might cause the PIC to stutter and miss a half-cycle or more.
>> (The PIC has to actively turn the triac on for each half-cycle.)
>> Earlier Smarthome designs (e.g. 1132B and 2000STW) have
>> been overly sensitive to disruption by out of band noise.
>
>I believe Dan had also mentioned this unwanted PIC interaction as a possible
>cause for the flickering.  FWIW, this seems to be the likely explanation
>because reports indicate the V2's are emitting EMI that's extending well
>beyond the local circuit.  Something's leaking out of the V2's that's strong
>enough to fool nearby devices into behaving erratically.  There's not much
>other than EMI and voltage fluctuations to explain flickering.   I believe I
>read that the EMI placed on the line by a triac is 90 degrees out of phase
>with the original AC cycle.  Is that what's causing nearby dimmers to
>flicker?
>
>I also read that chokes are selected for dimmer circuits that limit the rate
>of rise of current to minimize EMI and that the choke by itself does not
>solve the noise problem.  The self-capacitance of the inductor means they
>typically resonate below 200 kHz and look like capacitors to disturbances
>above the resonance frequency.
>
>(BTW, the odd effect I saw with my dad's shoplite flickering was caused by a
>bulb that was on its last legs.  Even the small voltage drop caused by the
>charger was enough to destabilize the lamp, which is electronically
>ballasted.)
>
>It seem as if the flicker issue is pretty close to resolved.  Now, what
>about the flashing/blinking reports?  Would triac noise be able to trigger a
>very brief on/off cycle?  It seems to get back to the issue of whether noise
>on the line can cause unwanted switch operation.  IIRC, a while back you had
>discovered that a big enough spike will cause X-10 switches to cycle without
>valid input.  Is that what we're seeing here?
>
>One more stupid question:  Are triacs and thyristors the same thing?