Silicon Valley Redneck

And another!

Another purple envelope arrived today, but that's not really the topic (that'n is a little-bitty 4-layer design for playing with in background mode: a range-boosted 802.15.4 radio module).

No, today's topic is a board that came in a couple of weeks ago, and I just got it (mostly) working, with a couple of patch wires because Careless*.  Still needs a barnacle tacked on, because I changed the design after the board was out to fab.  Also needs AVR firmware installed (probably Arduino bootloader plus a small sketch), and a connector that should be arriving Wednesday-ish.

It's my first home-baked Raspberry Pi Crust!  I'm being coy about its function for the moment, as this is meant to become a product, but it's a cute little rascal, and I should be making connections with a corner of the prospective market at the Maker Faire this coming weekend.

There's also a Beagle Cape version in the works, along with a tidied-up respin of the Pi Crust, in the next month or so.

* When you're knocking out board designs in an afternoon, and not on a deadline, and prototype fab cost is under $10, convening a $2500 formal design review Just Ain't Happenin'.  Also, those design reviews never seem to catch the sorts of errors I had on this board anyway.

What a revolting development.

Well, I got the USB end of the Probe-O-Zap all built, and set up so's It Oughtta Work, and....

And, while the FT201X chip has been known to work with my office workstation, the lab Linux machine happily enumerates it but doesn't recognize it as a serial port.

Because the lab machine is running kernel 2.6.32 is why.

Because it's still on Debian Squeeze, as was the shiny new Stable version a couple years back.

And that old kernel had never heard of an FT201X, what with the part just having come out last year.

And so it's time for the ol' apt-get dist-upgrade, and chewing up vast amounts of bandwidth the next few hours.

In other Probe-O-Zap-related news, I've been pondering how to package that little board and its Mystery Component, and had pretty much settled on sticking it in a mold and pouring in brightly-colored polyurethane potting compound.  Hmmm, the mold could perhaps be made of Teflon, with a nice slick hole bored in it.  Lessee, it needs a 1/2"-ish cylindrical-ish section, then a tapered step down, then a 10mm-ish bit.  Be nice if the cylindrical stretch, being a couple inches long, had a bit of a draft angle.  Now, where might I obtain a suitably-profiled reamer for the task, and what'll it cost to have it made?  Aha!  Inspiration strikes!  The profile I'm looking for bears a striking resemblance to (digs out old reference book) a .375 Weatherby cartridge.  So, all's I need is a possibly-stock chambering reamer!

(Brownells has one in stock for .375 H&H, but the shoulder diameter is too small.  Unless I taper the PC board, which is plausible.  Clymer doesn't seem to do .375 Weatherby.  Oh, well: board, encapsulation, and caliber choice are all subject to further negotiation.)

Update: Midway turns out to sell .375 Weatherby reamers (with a 30-day lead time).  Choice of slightly-undersized sizer-die reamer or exact-sized chamber finishing reamer.  I assume either of these will cut Teflon.  And, given that they're made for cutting tough grades of steel, probably last for a great many mold cavities in Teflon.

Volume

So: I look at the planned respin of the Probe-O-Zap, and figure I can shoehorn the revised & updated (and slightly more complex) design into 1.875 x 0.475 inches.

Now, how about a quick online quote for, say, 100 pieces, with a not-especially-fast turn?  Preferably panelized for more efficient assembly?

Hmph.  Even without panelization, Sunstone wants several dollars per board, in hundreds.  Sierra Circuits and PC Fab Express, at least using their easy-quote form, won't touch anything under 0.5" wide, and with that adjustment they're still only a little cheaper than Sunstone.

Given that a pile'o'boards on a 2-week-ish turn from OSH Park would run about $1.50 per board, with easy ordering... maybe I'll get in touch and see about quantity orders, and whether they can be delivered in panelized form.  And also talk to an NPI assembly house over in Fremont, that might have a favorite fab service.

Or, I could try the usual vendors' full-service options.  Which I've always assumed were even more expensive, but maybe it gets better in hundreds.  Or get in touch with an overseas fab.

Not that I'm making commitments at this point, not yet having done a market survey nor secured a steady supply of the Mystery Component.  My supplier claims to have at least hundreds available at a good price, but a single run of 100 units probably isn't worth the trouble (best guesstimate, it starts making sense at around 1000 pieces in the first year).

Update: (a) yes, OSH Park can do what I want; (b) a regular PCB manufacturer can do the same thing, cheaper, if I'm willing to order maybe 1000 pieces at a whack, which is not the case at this point, but perhaps in a few months, if the component situation proves favorable and the product really takes off.  Near-term, likely case is 10 panelettes with 20 little boards per.

Got another purple envelope today

Containing three of my latest purple board - yes, the one that goes with the mysterious East European item - and a purple sticker.

Update: assembled the right (i.e., Eastern-facing) end of the board this afternoon, and spent an hour or so playing mad scientist with this, various bench instruments (power supply, pulse generator, scope, voltmeter), and the Mystery Item.  It works!  Next up: populate the MCU and a few other parts, and throw together the basic operating firmware.  Then, the left end (USB interface), and some rinky-dink Ruby program for demo purposes.

I expect to be turning this into a Genuine Product (actually, 2 or 3 closely-related products) sort of June-July-ish, assuming an ongoing supply of Mystery Items can be negotiated.  Watch for an announcement.

Update, Sunday afternoon: MCU is now controlling the interface section (in a rudimentary way, as I'm missing a small but important component at the moment), and taking measurements.  Another few hours should have the dingus fully operational.  When those few hours will be... well, I've got a busy week coming up.

Update, Thursday lunchtime: hardware is now fully armed and operational.  I'm sending measurements back to the host PC via the FT201X.  Just need to implement the rest of the firmware features, and it'll look remarkably like a Product (modulo packaging and suchlike details).  This afternoon, though, there's Real Work to be done.

Egad, even the envelopes are purple now!

Well, you can tell at a glance when your prototypes arrive from OSH Park:

Yes, the tiny out-of-focus thing is an actual prototype of an actual board with (assuming it works as expected) a useful function.  The big one is a crust for the Raspberry Pi.

Update, Thursday afternoon: I'm informed that the next prototype is in the mail.  Don't know when the critical part for that one will be arriving via slow boat from Ukraine....

Update, Friday morning: suspicious parcel with Cyrillic writing on it arrived.  Return address is in Odessa, obviously cause for suspicion.  Anyway, critical long-lead-time parts are here, hurray!  ...As is the latest shipment from Digi-Key, with the bits I'd forgotten on the previous order...

Anyway, this all points to a happy busy weekend of building & testing.

New toy

My Beaglebone Black just arrived!

Cute li'l rascal.  And, though it sports a micro-SD socket, the OS lives in 2GB of on-board flash.  So, the add-on memory card is optional.  Looks like a good configuration for designing into custom programmable instruments, with the power, Ethernet, and USB connectors exposed to the world.

Haven't powered it up yet.  Got work to do, and here it is the end of the month, Walpurgisnacht, and all that.

But!  I now have the mechanical sample that allows me to make sense of the cape template I found in someone-or-other's library (prolly Lady Ada's).  Looks like I pretty much have to keep my tall components - in particular, things that go at the ends, such as connectors - topside, unlike the RasPi crust.

For now, back to work.  Assuming, of course, that I can get the darn furry Huckle-beastie to go take a nap someplace that isn't on top of me.

OSH Park is a disruptive business.

Note how many PCB designs I've sent out to fab in the past couple of weeks, Just Because.

Latest one went out a few minutes ago.  It's a 4-layer design, for technical reasons, so I don't expect to get my boards for about a month and a half, but consider:

If someone were paying me to develop this critter, and needed it in a hurry, I'd fork over the $250 (or whatever) of the client's money to have Sierra Circuits turn the boards in a week.  Or even more money, for a shorter turn.

Otherwise, and if OSH Park weren't there... well, it pretty much Wouldn't Happen.  Oh, I might use BatchPCB, or even one of those cheap-ish but epically inconvenient overseas services that'll do small lots of little 4-layer boards, but I'd have to give serious consideration to the cost and effort.

As it is, I think: 4 layers, area a bit under a square inch, so a bit under ten bucks to have a batch made.  And I sign in, and upload my EAGLE file, and the auto-quoter quotes me $9.60, and the auto-previewer shows me pictures that look like my board, and then comes the minor inconvenience of having to go through PayPal (or Google Checkout) to pay for it.

So here I am, busily prototyping things that may or may not become products, just because I have the tools and the ideas and there's a cheap and easy fab service for prototyping small boards.

And, no, they won't be taking very much business away from the quick-turn and full-service fab houses.  This is entirely new business that, if the new service weren't there, wouldn't be happening at all.

(And a lot of companies aren't into the Efficient Prototyping thing at all, and insist on going the slow-turn, expensive, full-service route for first prototypes.  Like, an 8-week, multi-thousand-dollar fab operation, plus a full set of release documents, for a board that we hope will be good enough to deliver to end customers... when Sierra could have made a set of good-enough-for-engineering boards in under a week for maybe $350, and we could have used the next week making sure the hardware design was OK, then overlapped 8 weeks of serious testing with the production fab run.  Yes, I've seen this in more than one place.)

Another great idea down the drain...

...well, part of one, anyway.  The part I just spent a couple of hours on.

The big idea: nifty gadget (purpose undisclosed at this point, though it's a sensor of some sort) with a little AVR handing some basic control and monitoring functions, and a USB connection for power and communication.

The great little idea: hey, why not use at ATmega8U2 (or 16U2) and have the USB host comm stuff right there on the chip?  Could save maybe a buck, a couple of components, and half an inch of board length.

Looks plausible, initially, so I spend a couple of hours revising the board design, getting the layout just so, and...

Wait.  Before I send this thing out to fab, maybe... just maybe... I should make sure the firmware support is really there.

Well, yes, it is.  Sort of.  But, before I even get into digging into the licensing terms, I'm reminded of a little something.

VID/PID.  Gots to haz them.  Even if the device conforms to a defined class, and should be handled generically by a class driver.  Well... Linux is forgiving in this regard... but Windows, on encountering a CDC-ACM sort of com-port device, refuses to recognize it without at least a .INF file, which is identified in terms of VID/PID.

And, since I'm not about to pay the dues to become one of the 65534 (or whatever) possible vendors, this is kind of Out.

And so I return to Plan A, in which I use an ATmega88 or some such, plus an FT201XQ, which brings its own VID/PID and works with the FTDI virtual COM port driver.

Drives the price of the gadget back up a tad, and puts that quarter-square-inch back in, but I wont need to muck about with the USB nonsense.

Update: aaaand... the board with the ATmega88+FT201XQ is out to fab.  Another of my 1-square-inch wonders.  Fiberglass should turn up in my mailbox in a bit over 2 weeks, which is about the earliest I'll be ready to do anything with it anyway.

Took the afternoon off...

...put on my second-loudest Hawaiian shirt, and spent a few hours wandering around Design West, down at the San Jose Convention Center.

With the Embedded Systems Conference having been merged with various other trade shows, the result is enough to justify the drive and parking, and even a moderate wander around the expo floor took all afternoon.

Lotsa new shinies on display, and I picked up some useful information from vendors (even if the information was "nobody makes those in the temperature range you need").

I didn't engage in organized schwag-hunting, and won't be returning tomorrow to try to win a new-generation BeagleBone at the TI booth.  (Having heard from the BeagleBone folks that they'd actually shipped the first batch to distributors this morning, I ordered one from Mouser when I got home.)

I won a drawing!  I stopped by the EarthLCD booth to say hi, and was persuaded to drop a business card in the jar.  Returning at 1400 sharp, i.e., 5 minutes later, I heard several names go unanswered, one after another... and then my name.  Must be present to win and all that, and now I've got this ARLCD thingie, which seems to be a combination ezLCD and Arduino Uno.  Have to give it a try, in the not-too-distant future.  Might see what's involved in throwing together a cute li'l graphical status display, then feed it data from the RasPi with the new crust attached (once the crust is here and assembled).

Technical note

If you're trying to build a flyback converter, in a tiny space, to provide up to 500V at negligible current, for whatever reason, the TDK ATB322524-0110 seems a handy little off-the-shelf transformer for the purpose.

But!  It does need a low-on-resistance FET.  I've just been trying to drive one with a 2N7002BK, and even at 5V supply and 5V gate drive, I can't coax more'n 200V out of that puppy.  Around 1.7µs on-time, the transistor just doesn't maintain full voltage across the primary anymore.

I'll try again with a BSS138K after my next Digi-Key order.  In the board respin, I'll be changing to a different footprint to accommodate something with even better on-resistance.

And, yes, I currently have two reasons for wanting to step small-battery power up to a few hundred Volts in a small space.  Both are currently secret, because business.

Update: substituting the BSS138K, I get right about 400V out at 5V supply, 2.1 µs pulse width, 50 KHz.  Current draw is about 60 mA.  Looking at the drain waveform, the FET's resistance is still a significant factor.  The new board (currently out to fab) uses a still-lower-on-resistance part with a somewhat larger footprint, and should be more efficient.  Also, given that the load is basically zero, dropping the frequency to 25 KHz and bumping the pulse width to 2.5 µs keeps the output voltage up while usefully decreasing current draw.

In the live application, I can tolerate a fair amount of output ripple (more in one application than the other), so the MCU code to manage the PWM will probably be skipping a lot of cycles, and just turning the drive back on when the HV falls out of tolerance.