So far the only soldering I've finished today is fixing a Blue Snowflake I got as a gift. Which, I am sorry to have to say, has some piss-poor engineering inside it. I understand necessary compromise, and how process will make unforeseen changes, but who would feel good signing off on this one!
Here's the problem. Daughterboard holding the mini USB connector. That makes sense; put the stress there, use that to screw to the chassis, have the electronics snug into it. The main circuit board, though, is suspended on two 3-pin headers. Which means it wiggles, and you get poor USB connectivity. After I'd reheated all the solder joints from USB jack to main circuit board, I re-assembled, and then put a couple drops of hot melt glue in to keep the thing from rocking back and forth on the connectors and breaking connection.
The boards are notched to go around the swivel nut that connects chassis to frame. And somehow, the USB was put in at about a 17-degree angle to the notch. Which means, when the board is assembled and installed, the swivel nut misses the notch entirely, and is all but jammed into the connector to the actual mic capsule. It makes it extremely difficult to finish the assembly without knocking out the connector (or tapping the circuit board off it's precarious balance on its headers.) Broken? I'm surprised the thing actually survived being shipped in working condition!
A vaguely similar problem besets my Sennheiser belt packs. The pack usually ends up strapped close to an actor's body, often in the natural curve of the small of the back. The microphone element is connected to the transmitter with a 1/8th TRS locking connector (expensive little things; I'm paying almost $12 each). The length of the connector means lots of leverage on the jack, causing several problems; the connectors unscrew, the wires break inside the plug, and the jack itself snaps off inside.
The latter is a pain because the circuit board is crammed with surface-mount components. The last one, I repaired by carefully cutting the broken jack to ribbons and then individually de-soldering the remaining scraps of the legs. The one previous, I tore the traces off the board trying to lever it out, and that means that whole unit has to be shipping back to Sennheiser for repair.
Despite these directly job-related repairs, I also have many of the parts from my Duck Node experiments spread out on the work bench as well. I want to have something ready for whatever comes up on the next production, even if it is another hand-soldered prototype instead of a proper board. Tangentially related to this, I'm going to be doing a soldering project with my youngest niece this Friday; a 4x4x4 LED grid from Seeedstudio. Which is also intended as a gateway drug into programming.
From the top, this is my wireless EasyButton (which may or may not currently have a configured XBee in it -- I keep switching the things around). Then an Arduino Trinket from Adafruit; a more-or-less Arduino-compatible board based around an ATtiny. I've been thinking about it for the DuckNode because it offers built-in USB programming for under $10. But it is also severely limited in the number of I/O pins, plus the USB bootstrapper is fragile (hence the festoon of wires on it now; those are programming leads). So bumping up to over $20 for an Arduino Mini probably makes more sense.
In the lower right is an abandoned proof-of-concept for an ATtiny-driven 3W RGB. It is being switched through power darlingtons, with hefty (1 watt) ballast resistors. Why I wanted to go for the more efficient constant-current drivers. And on the left is a proof-of-concept for the DuckNode concept, along with the wrist holster; an accelerometer paired with an XBee. For my tests, I parsed the output of the accelerometer in a Processing application, and fired off a sound cue when it crossed the right threshold.
The problem here...and why the project got suddenly so much more complicated...is very few accelerometers will output in analog that can be easily patched through to the ADCs on an XBee. I haven't seen a single one of the integrated accelerometer/compass units that wasn't digital (usually I2C). And that is such a pain to interface with XBee you are better off parsing through a microprocessor and handing that data off in a serial stream to the XBee. Which is also more power-efficient.
And, yeah; XBees are expensive, have a large footprint (relative to some other radios), and they (like a lot of electronics) are moving...from the Series 1 I own, to Series 2 -- both of which are deprecated in favor of the (much less useful to me) Zigbee protocol. But all the other options take time to learn and develop around. I already know the XBees. And I've used them in real stage conditions.
Sigh. And I also have to cut down the cords (they are too long and thin to be safe to use) on some chorus microphones before Saturday, where I'll be tacking them to the unistrut with zip ties. I filled several pages of notebook trying to come up with a combined unistrut bolt and shock-mount mic clamp for them that I could machine up myself, but no luck. Which means my first milling project may indeed be the Jubal Early.
Which is also going slowly. I'm finally getting a grip on how the parts fit together, but that meant throwing away the first attempt at a poly model. Arg. This would make so much more sense in a proper CAD, or at least in a modeling application that was more parametric. Carrara is not a NURBS modeler. What you build is what you get (aside from the option of tessellation -- but that tool has extremely limited options and tends to make a mess of models).
I really, really need to get comfortable with Blender, but out of all of the things I'm constantly having to learn and re-learn, software is the slowest and most frustrating of all of them.
No comments:
Post a Comment