How to not build a dimmer pack

While diagnosing a dimmer pack that had been misbehaving somewhat during a show, I took the opportunity for a peek at the innards. This is a Chauvet PRO-D6, a slightly higher-end six-channel pack than your typical single-input 4-channel units found in many small DJ rigs, but still more or less aimed at the DJ market. It's likely intended to compete at a lower price point against the Leprecon LD-360 and similar tree-mount units.
Inside the Chauvet PRO-D6 Inside the Chauvet PRO-D6
It's the typical dual-input-plug design, which splits two 20A feeds to three dimmer outputs each. There's a smallish toroid inductor on each output and a bit of token filtering on the inputs, all told probably just enough to take off the really sharp harmonics but not really suppress noise. The input cordsets are 12/3 SJTW with *very* stiff insulation, typical of many Chinese power cords these days. The molded plugs are 20 amp 5-20P type with the sideways neutral blade, sometimes making it a little harder to figure out where to plug them in. The ironic thing about most 15A vs. 20A Edison outlets is that the internal brass is by and large exactly the same, the faceplate just has the "tee" slot in the neutral hole or not.

The pack is not really designed to be serviceable, and is in fact hard to gain reasonable access inside because all the wiring is crammed in and closely bound up with zip-ties. I've got it pushed apart about as far as I dare in the lefthand shot. When the outlet panel is in place on the extruded housing, the wiring is all mashed together and it's actually hard to fully seat the panel. Close examination [click each pic for a larger version] shows a few places where the insulation has been pinched and nicked by the surrounding parts. There's a fair bit of empty space near the ends of the unit, and it would have been easy to use a little more wire and make a nice service-loop out of the six output leads, four neutrals and a ground to allow the panel to swing conveniently away.

This also shows a poor choice for output receptacles, using the "back-stab" type of spring-lock connection into the sockets instead of solid lugs and screws. Ask any electrician about the long-term viability of such back-connect duplex outlets, especially under sustained high loads. The present-day answer is to avoid them like the plague. At least the neutrals are doubled up and run to either end of the chain across the outlets, to avoid loading any one connection too hard. Is that all solid-core wire in there? It's quite stiff but I couldn't really tell, and didn't want to tug on any of it too hard.

At this point, I gave up and put the thing back together.

But what was the original observed problem? When trying to bring lights up slowly, anything connected to the second half of the pack in channels 4 - 6 would not come up smoothly, but suddenly pop on somewhere over halfway as soon as any nonzero level command was sent via DMX. The same thing would happen when the pack's control panel was used to set a manual level -- as soon as anything above 0% was requested, the high channels would be a-burnin' brightly away while the low ones would just barely begin to warm the filaments. Later, however, I couldn't reproduce this at home with the two power inputs plugged into the same circuit, or even two different circuits on different legs of the house panel. At that point it began to dawn on me what was going on, and a quick dig into the PDF manual for the unit revealed that it's a documented issue:

     "The power for the LINE A and the LINE B circuits must share the same
     electrical phase.  The unit will not operate properly if they are on
     different phases."
At the venue we were on three-phase power, and the two inputs were on circuits 120 degrees apart. Whereas at home, I only have circuits either 0 or 180 degrees apart. Apparently the triac firing time reference is derived only from the first circuit, which places it at the wrong part of the phase for the second circuit if its zero-crossing doesn't line up. So the high half of the pack was working, just being triggered at the wrong time by the reference from the low half.

What this means is that to use these packs in most commercial venues and have smooth dimming, the two line inputs *have* to be plugged into two circuits on the same phase like the manual says. Sheez. My old Leprecon packs with their mostly analog innards don't even have a whisper of this problem; they have a second timing circuit designed to handle just such situations with jumpers to use it if the "High Power" option and second input is installed at manufacture time. What do you do if you're in that tiny basement black-box on the only two available convenience outlets and have no other choice about power??

To Chauvet's credit, the extruded housing is fairly nicely made and uses machine screws into tapped holes to hold the end plates on, as opposed to the self-tapping disasters seen in some other cheap construction. And they use 25-amp bta26-600b triacs, giving generous capacity overhead if one chooses to run a heavy load from one channel. The configuration and options accessible from the front panel are quite versatile, allowing various DMX-loss fallback modes, standalone sequence programming, and interesting setups with each individual dimmer at its own arbitrary DMX address if needed.

No dimmer is guaranteed to completely turn off an unloaded output, due to the nature of how the SCRs or triacs and related circuitry are wired. Usually a connected load will soak up any remaining trickle and bring an output to zero volts without drawing any meaningful power, so oddities in open-circuit conditions aren't something anyone really thinks about and safe practice is to unplug something being worked on anyway. With the Chauvet, even at 0% output, *full line voltage* is capacitively coupled to each output and there's enough reserve power in there to make a noticeable little spark -- enough to visibly blink a lamp filament, even -- as a load is connected if the line voltage happens to be near the crest at the same instant. It's also enough to give a painful little tingle to a human touching an unloaded output, indicating that the inherent zero-output impedance in the Chauvet pack might be a bit lower than safety-conscious design would suggest.

_H*   110215