Silicon Valley Redneck

Troubled sleep

I woke up around 0240 this morning, and for a change it wasn't Huckleberry's fault.

I was having a strange nightmare.  I was out on a hike, along (and in) a good-sized creek or itty-bitty river out in the boonies, on a fine late-summer California day, with a group.  It wasn't clear whether I was the instigator of this hike, or if the group had spontaneously assembled.  The other hikers were mostly a fair bit younger than I.

One thirtysomething guy, apparently in good shape and with good equipment, whose name might have been Kurt, had a mishap of some sort; the details were inconsistent even before I woke up, but I'm certain we heard a splash and didn't see him afterward.  We looked around.  Gradually, the group dissipated, leaving me and two women, one of whom had some interest in Kurt (it's not clear whether they'd arrived together or just met on the hike and promptly hit it off).  Then one of us spotted his pack at the bottom of the creek.  And his arm.  And... oh crap.

So I grabbed him by the pack and dragged him ashore.  Submerged for... maybe half an hour?  Long enough for people to lose interest and wander off, anyway.  And the water is far from cold.  Really-for-sure dead, right?

At this point, my personality disorder came to the fore.  What's the protocol for this*?  Should we be notifying someone?  Who?  How?  Where's the checklist?

Then I woke up, and wasn't about to get back to sleep real soon, so I read for a while, then tended the critters, and went back to bed.

 

I didn't get much sleep after returning to bed, either, because by that time Huckleberry was definitely awake and figured I should be too.  But I did drift off a few times, to strange dreams.

I was repairing my driveway.  Or having it repaired.  Or had just had it repaired.  Whatever: there was a dirt patch on one side, at the sidewalk end, about where there's currently a metal post that (before my time) once supported a chain-link gate.  Why there was a dirt patch there, I don't know, but it seemed to make sense in context, and I'd planted things in it and was disturbed to note that someone had come along and planted little wooden stakes with pink ribbons on them, with spritzes of pink spray paint for good measure: apparently markers in service of the city's plan to issue everyone curbside mailboxes (which plan, so far as I know, only exists within that dream).

Then there came to my attention pet toys that were basically tennis balls (non-squeaky) the size of ping-pong balls, for small dogs and possibly cats.  A stranger popped into the dream just long enough to tell me that he'd constructed an outsized revolver that fired those things, possibly using nail-gun blanks for propulsion - this wasn't clear, and wouldn't CO₂ be a better choice? - but I'm certain is was some sort of deflagratory gas source.

When I woke up from that one, Huckleberry was aggressively looking for mischief to get into, so I decided it must be morning.

 

And that, folks, is why I've been groggy and out of sorts all day.

 

* First, make sure he's dead.  Bang!  OK, now what?**

** Usually presented as a redneck joke, but I think I first heard it as a Norsky joke.

Ol' Combover says stuff

I tried to read this interview with Belloc's Bison.  Really, I tried.

When asked a question, he says stuff.  A bunch of stuff.  It's not in any way coherent.  It's just a bunch of stuff.

Maybe I'm missing something by reading the transcript instead of listening to His Voice?

It sure looks like he's just throwing stuff out there, and some of it sticks.  Has the political climate gotten so bad, and the electorate so accustomed to curate's-egg politics, that large numbers of voters will really latch on to tiny scraps of apparent sense (from any given voter's perspective) amid a stream of idiocy?

I'd already noticed that Mr. Super-Duper Businessman has a grasp of basic economic principles that's right down there with Bernie's, but I guess that shouldn't be a surprise; big business, and especially big business involving big-city real estate and casinos, is all about politics, and doesn't really call for even the most rudimentary understanding of economics.

It's another of Those Mornings, isn't it?

Wake up at approximately normal time... eventually wander into office, turn on monitor, push mouse, stumble into kitchen in search of caffeine, return to office...

... This computer has been locked by eric.  Password?

It's.  Not.  Accepting.  My.  Password.

Try toggling caps lock: that's not it.

Switch keyboard & monitor over to server.  See a flurry of error messages, apparently related to the interface (not, near as I can tell, the surface) of the second disk drive.

Um.

Reboot the server into single-user mode.  Dismount /second and force a check of /dev/sdb1.  Clean, but: as e2fsck is finishing up, I get a warning about CPU temperature, and CPU speed being throttled.

Then I have to wait for the workstation to get its NFS client status sorted out, which it does eventually (restarting nfs-common may or may not have helped).

So I guess after breakfast, and before laundry and Family Stuff and whatnot, I'd better shut down both office machines and remove the cat hair.

Update: Wasn't so much cat hair as regular old household dust.  There was a lot of it.

I think the CPU temperature reading is exaggerated, though; it was showing unreasonably high when I turned the machine back on after several minutes and went directly to the BIOS status screen.  Or maybe it's just going directly to the temperature at which the fan runs at full speed, and stabilizing there?

Anyway, things seem to be back to normal.  The SMART statistics for the drive in question are totally healthy, and I haven't seen another CPU temperature warning (yet).

So it must be time to gather up laundry and a book, and go forth into the day.

I wish they wouldn't do that.

Being brain-fried and crosseyed from attempting to clarify some ought-to-be-simple documentation for a client from years gone by, I take a break and try to do something else that ought to be simple.

I take another look at the Linux USB test-and-measurement class thing.

It appears that there are two kernel drivers for it: one that comes standard with any recent version of Linux, and one from Agilent (or whatever they're called this week).

They're both called usbtmc, but they don't, quite, do the same thing.  Like, apparently the one from Agilent creates /dev/usbtmc0 as a meta-device returning a list of connected devices, while the default driver normally puts the first TMC device it sees at usbtmc0.

And yet the two drivers have the same name, and use similar-looking device names.  So any application that uses TMC via the driver (rather than, say, libusb) must either figure out which driver is in use, or insist on having a particular driver configured.

I like the Agilent approach... but how come it and the default driver have the same name?  (And maybe the directory meta-device should have a name other than usbtmc0, eh?  Or would that mess with the scheme whereby /dev is populated?  How about /dev/bus/usbtmc/*?)

Also, there's some oddity with how they behave, and it looks like using a terminal app to try doing *IDN? just to see if the device is alive doesn't work... probably because terminal apps try to set the Baud rate, which virtual-com-port devices (or their drivers) will typically accept even if there's no Baud rate to be set.  Too bad the TMC driver doesn't do that, just to be friendly.

Update: The other oddity is of course the "connection timed out" error.  I'm not clear on whether this is an issue with both drivers, or just the default (non-Agilent) one.  It happens with cat, and it happens with dd, and it makes life complicated for those who are just trying to do a quick-and-dirty communication check (I haven't tried talking to the instrument using Ruby yet).

With a bit of trickery, though, I can establish that it is communicating:

eric@magrat:~$ echo '*IDN?' > /dev/usbtmc2; dd if=/dev/usbtmc2 of=foo.out bs=1
dd: reading `/dev/usbtmc2': Connection timed out
52+0 records in
52+0 records out
52 bytes (52 B) copied, 5.05962 s, 0.0 kB/s
eric@magrat:~$ cat foo.out
RIGOL TECHNOLOGIES,MSO1104Z,DS1ZD171600408,00.04.03
eric@magrat:~$

For some reason, the workstation puts the first instrument at usbtmc2.  I have no clue why it does this.

This isn't really a case of my wanting to talk to the new scope from the computer right away; the scope is just the only TMC device I have handy, and I'm trying to use it as a reference as I ponder implementing my own TMC device protocol layer for my own purposes.  (And then I'll maybe release the project, for those interested in turning a FRDM board into a TMC device.  When I get it done.  Which won't be right away, because when I do get my brain back in gear, it needs to be working on stuff for clients.)

Wakey wakey

I mentioned getting home late last night?

This morning the landline rang at 0340.  Mnfxt?  Gotta be either a parental emergency or a wrong number.  Better answer it.

It's GrannyButton Industries.  They got a signal from my mother's GrannyButton, didn't get a response to attempts to contact her, and sent someone out to check, but he can't tell me anything more because HIPAA.

Which is weird, because the other times GBI has called me, there's been further information, like "everything's OK" or "Stanford Hospital" or some such.

So I try their landline, and get their answering machine.  Try my father's cellphone, and it goes direct to voicemail: he doesn't have it turned on (this being the normal condition).

Make sure all critters have food, water, clean litterboxes, and whatnot, then try the phones again: same.

Blrg?  Get dressed and get in the car.  At least there's basically no traffic this time of the morning.  Arrive at their house: both cars in the driveway, no lights on, all appears normal.  Nobody answers the phone, nor the doorbell.  Well, the dog gives a sort of halfhearted answer to the doorbell.  All is exactly as I would expect if I showed up there at this hour.

Ponder.  Try phone and doorbell again.  If the GBI Response Team didn't get a response, they wouldn't break the door down, would they?  Maybe I'd better knock on their bedroom door?

Dig out the key; let myself in... by this time I've made enough commotion that my father comes stumbling out in his bathrobe.

Turns out they hadn't answered the call from GBI because they'd been asleep, and nothing was wrong in the first place.  The fire department had come out, shouted through an open window, gotten their attention, determined that all was well, and left.

And they hadn't answer the phone when I called because, by the time GBI called me, the fire department had been and gone and they'd gone back to sleep already.

So I guess now it's time for breakfast, before taking the car in for scheduled maintenance.

Meanwhile, Joy found it necessary to reformat her Windows 7 laptop, and re-install everything, and re-do all the updates... and it sounds like it's stuck in a "Windows is installing updates" loop.  Wonder if it tried to update itself to WinX, automagically and without asking, and botched it?

Update: Dropped the car off at 0650, before the morning traffic got started.  Being bright & early, I had a very short wait for the shuttle ride home.  About 0910, I got the call: car's done; everything looks good.

Also, I finished up the dist-upgrade (also doing the autoremove) on both the lab machine and the laptop.  Then I plugged the finicky little inkjet printer into the laptop.  Nothing.  Check dmesg.  Hmmm... what we have here is a failure to enumerate.  Happens sometimes in the wonderful world of USB.  Try a different port?  Aha!  I promptly get a pop-up announcement that a printer has been found and configured.  Was it ever this easy in Windows?  Try a test page: looks fine.

Also, the upgrade let me configure things such that users in group wireshark can do packet capture without having to become root.  This Just Works (once I added myself to the group in question, then logged off and back on).

Now I'd better get some real work done.  Then retrieve the car during some low-traffic time.

Something in the air?

I spent most of yesterday being miserable and unproductive, with a nasty headache, irritated nose, and sometimes irritated eyes.  I assumed that these resulted from some combination of caffeine, blood sugar, stress, and overly-dry indoor air.

I woke up this morning with the same crud, but went for morning walkies anyway.  Up in the hills, I felt fine.  Back in the flatlands, though, before I'd even gotten home the symptoms had returned.

Hmmm.  Well, traffic is bad, there's currently no wind whatsoever, I don't recall there having been any wind yesterday, and the last rain was several days ago.  Could be we've got a serious case of smog buildup.

This is not good.  I've got a couple of urgent tasks for clients, and stuff I want to get on this month's timesheet, and here I am needing a long nap or an all-day hike in the hills.  Most of the work really has to be done at my desk, not in a park.

Update: Moderately better this evening.  There's supposed to be some rain tonight, so if the problem is something in the air, maybe it'll get washed away.

Got back kind of late from an unexpected (but interesting and productive) evening meeting.

The only thing I actually accomplished today that was even remotely on my checklist was to update my laptop to the latest version of Debian.  That took a long time, probably because there's a lot of software installed on it, which made for a humongous download to update it all.  It seems to be OK, and EAGLE v7 appears to behave itself, which suggests that I probably do want to update the workstation at the earliest opportunity.  Not this weekend, though; maybe next.  Gotta do some more testing with the laptop, just to make sure nothing important got broken.

Catching-up-on-work stuff: Saturday is Family Stuff, so I guess I'd better be productive tomorrow and Sunday.  Gah.

Is this National Shoulder-Cat Day?

Out walkies this morning, I stopped to say hello to Sprite, the Dread Guardian of the Farm, who (as she's sometimes done in the past) decided to climb up on my shoulder for a while.

Later, at home, Southmoon (after much sniffing and wondering whose scent was on my collar) also decided it was shoulder time, eventually finding her way to something approximating a proper position except that she was facing aft.  Then she decided to use her new vantage point to try to reach things up high.

This brings up a Really Important Issue that's been floating around my head for quite some time now: the awkward realities of having a cat riding around on one's shoulder.

It really is awkward.  For starters, even a small cat outweighs a large parrot (even a bloody great hyacinth macaw comes in under 4 pounds), so there's the weight to be supported.  There are four feet to be put somewhere.  And, if the shoulder has a head associated with it, that head ends up offset to make room for the passenger.

The whole exercise is very bad for the posture.  Really.  It it brings up the perennial question, where does Nimitz put his feet, anyway?

We have only partial specifications available for a Nimitz-class treecat and a Harrington-class treecat carrier.  In both cases, we're given length and displacement, but not beam, so it's necessary to make some wild approximations.

I'll assume that, while Harrington is quite a bit taller than I am, she's no wider across the shoulders (else she'd look a right freak), but her head is probably a bit narrower, so the available shoulder space may be about comparable to mine.

Nimitz's hull length and displacement just about match Huckleberry, but with 50% more legs to worry about.  Since a four-legged cat is a more awkward shoulder ornament than a two-legged parrot, we may assume that adding a further pair of legs will only complicate matters.

Now, Southmoon is positively dainty compared to Huckleberry, but with her on my port shoulder I was listing to starboard and forward to accommodate her and maintain balance - and having to move my head out of the space she was taking up.

So... where does Nimitz put his feet?  And his body?  Does Harrington have a permanent stoop and a neck kink?

Or is this the Wrong Kind Of Question, like trying to run the numbers on energy budgets for sci-fi spaceships?

OK, that was confusing for a while.

At least partly because my brain was fuzzy to begin with, which is why I was working on the Gadget instead of on something for a client, but anyway:

I finally hooked up the SCPI parser (and associated support functions) to the FreeRTOS main program, and imported a slightly abbreviated set of command tables from an old application on an entirely different platform.

Tried sending a command: nothing happened.

?

Step through the code.  The number of commands in the top-level table is coming up zero, instead of several.

??

Dig through the definitions, the map, the linker script, the... hey, wait.  The map shows where the table should be, but it doesn't show anything going into it.

???

Eventually, I find the problem!  The default project configuration tells the linker to optimize out unused sections.  This is a wondrous fine feature that (as I'd found, to my great confusion, in debugging another project) will go so far as to remove unused functions that came from the same source file as used functions.

Turning this option off makes the program work, but just about doubles the size.

So I put in some bogus code to "use" the tables (the parser finds them using linker-defined symbols).

Next thing will be to see if, the global section having been made to appear, my trick of appending to the array from a variety of source files (using the linker to combine their contributions to the section) will work without having the stupid bogus reference in each source file that makes a contribution.

This also has Implications for my idea of porting AGROS and using Freescale's hardware abstraction layer; AGROS makes extensive use of magic sections that get assembled by the linker and contain no symbols referenced directly by anything.  If I have to disable that optimization, how much unwanted code will get dragged in?

Life imitates Futurama!

Remember how Bender competed in the 3004 Olympics as a fembot?

Well, it looks like the IOC has almost caught up with that!

Anatomically male athletes who've spent their growing, muscle- and bone-developing years awash in testosterone will now be allowed to compete as women as long as they drain their testosteroil and replace it with femzoil.

They will, however, be required to have said oil change performed a year before the competition, a sad sign of how far we have yet to go.

...Wait.  Will those East German women be subject to retroactive testosterone-level checks?

My AVR assembly is rusty

And I'm both lazy and overcaffeinated at the moment, never a good combination.

So I wrote a stupid little application in C.  It compiles to 194 bytes, and seems to run just fine on the ATtiny9 (in the hulking SOT23-6 package).  It does what I want, and doesn't involve three one-shots and a few gates.

I think the tiny9, with its 1K of program memory, may be overkill.  I could use a tiny4, and save, um, thruppence.  Well, maybe I'll stick with the more capable chip, just in case I come up with some more sophisticated use for such a thing.

So what is it?  Since I have no expectation of going commercial with it, I suppose I can reveal the awesome truth.  It's a reset-button controller for a bigger MCU.  Ya know how some of the nice new MCUs have ROM bootloaders, and there's some pin you pull low during RESET to get the bootloader instead of the application?  Well...

This 6-pin wonder actually uses 3 I/O pins: button in, reset out, and bootmode out.  Tap out Morse code on the button: the chip does nothing (I'm assuming for now that the button may also be routed to some other MCU input pin for application purposes; I may decide to do a passthrough to the bootmode output).  Hold the button for 4 seconds, and the reset output gets pulled low, with bootmode floating.  Continue holding the button, and at the 6.5 second mark the bootmode output gets pulled low.  Release the button, and reset is released; if bootmode is active, it'll be held low for a further 0.5 second before everything returns to normal.

When the button is not pressed, the tiny9 is in power-down mode, and ought to be drawing some fraction of a microAmp.

And for feedback, so the user knows when he's held the button long enough?  Hmmm.  There are dual LEDs available in 0603 packages (maybe smaller, too, but for hand-built prototypes I don't like really tiny LEDs).  Easy enough to make it light up red for reset, mixed for reset-to-booloader, and green for the post-reset interval while the bootmode line is still being driven.

And it saves a jumper!  (That, and it lets the button do double duty as hard reset and application-defined input.)

A minor inspiration

Been pondering property labels for my equipment, especially now that I have the shiny new portable scope, and the old portable is an occasional loaner.

Such labels are available with bar codes, typically encoding a serial number.

Hmmmmm....

How about a QR code (now that everyone's phone can read them), encoding a URL.

Point a browser at the URL, and, if you're authenticated, you get the property-management page for that item.  If you're not authenticated, you get the "if found, please return to..." page, maybe with some basic information on what the thing is.

Not that I'll be implementing this any time soon, so feel free.

Put away that tinfoil, lads.

Well, part of it, anyway.

Spotted here:

THERE are countless sites around the world deemed so classified that they are pixelated or blacked out by Google Earth.

Some are understandable, like some military sites, royal residences and say, the whole of North Korea. Others may surprise you. Pretty much all of them are the subject of conspiracy theories.

Wait, North Korea is pixelated?  How does the government of North Korea have that kind of influence with Google?  This does not make sense.

So I launch Google Earth myself, and look at a few spots in NK.  Not pixelated, but not exactly high-res either.  Kind of like some of the less fashionable areas of the U.S.

So, why?

Duh.  Base image quality is based on public, non-secret satellite imagery, which is kind of fuzzy, grayish, and washed-out.

For better imagery, Google has to pay for images from more modern satellites, or better still for aerial photos.

So when a vast area is fuzzy, grayish, and washed out... especially when that area has square corners... most likely Google didn't consider it worth paying extra for better imagery for that area.  And maybe aerial imagery wouldn't be available even if they did feel like paying for it.  It's not necessarily a decision to degrade available high-quality imagery.

Here's a bit of fun: point Google Earth at Mountain View, around, say, 101 and Rengstorff, just to pick a location at random.  Zoom, pan, tilt.  See all the effort they've put into it?

Now look at, say, Tower City, North Dakota.  Hm.  Well, they have been busy upgrading the imagery for flyover country - if memory serves, this used to be one of the fuzzy grayish areas - but they haven't put in nearly as much effort there as they have around their own neighborhood.  Finite resources, and all that.

There are still large areas within the U.S. with no street view, and I don't think it's lack of access - just a matter of the cars not having been through those places.

Oh, and that business with the Faroe Islands?  With the detail being "blurred" in sharp-edged strips?  If you look very closely, you'll notice that the strips are from different seasons.  Ergo, different vintage, and maybe different satellites.  Nothing to do with fishing rights.

While I'm not accomplishing things...

... near the top of my priority list, because those tend to require thought, and I'm still fuzzed out by this morning's FIEMI episode...

I did manage to get the ATtiny9 SOT-23-6 socket adapter wired up for connecting to the programming pod.  Now the paint markings (for the power-supply connections) need to dry.

Meanwhile, I've started the lab Linux box doing a dist-upgrade to the latest Debian.  If that goes well, I think the laptop will be next - or maybe the VPS.  Workstation is last.

Once the upgrade is done, I'll see if it plays nice with USB-TMC, or if I need to do more investigating on that front.

For now, off to read some more Cryptonomicon, or take a nap, whichever comes first.

Update: Aside from having to be done from the in-person console, taking a few hours (between downloading over not-the-fastest DSL and installing on not-the-fastest Celeron minibox), and needing manual intervention a few times during installation to confirm handling of non-default config files... that was fairly painless.  It's now running Debian 8, with KDE 4.14.2.  I just tried one of my Ruby / GTK+ scripts, and that seemed to work (I didn't have the actual hardware connected to verify functionality, but it woke up OK).

Looks like a fairly conservative upgrade (in keeping with Debian tradition), so no Qt5 and such, but I think it has the library versions needed for running the latest commercial software (EAGLE in particular).

Aside from checking out USB-TMC handling, I should see if the included version of CUPS/Foomatic/whatever plays nice with the label printer and the junior inkjet (both have been nominally supported since forever, but the former just doesn't work with the older CUPS, and the latter, when connected to my workstation, only gets halfway through a page before deciding the USB has stopped working).

Not doing any of that now, though.  It's way past Tinga's bedtime, and I've already disturbed her enough with going back into the lab repeatedly to check on progress.

Update 2: Setting up the label printer was pretty easy, and I can indeed print labels on it, though for some reason CUPS thinks they should be 100mm long - maybe I need to fiddle with the settings some more.  In general terms, it's working.  Haven't tried the inkjet nor the oscilloscope yet (and there are higher priorities today).

The operative word here is...

Over at The Atlantic, Joe Pinsker ponders why people buy milk, bread, and eggs before a winter storm, and notes (in more than one place):

But milk and eggs? Why, when the concern is that the power might go out, do people hoard things that need refrigeration, or even cooking?

Think, man!  Winter storm!  If the power goes out and the refrigerator starts warming up, simply move the perishables to an unheated spot.  Put 'em out on the porch, and they'll freeze.

If you have neither refrigeration nor cooking, we may assume your entire home is now an unheated spot, and the lack of refrigeration will shortly cease to be a problem.

Personally, I'd be more inclined to stock up on canned soup and Coleman fuel, but maybe that's just me.

FIEMI

Or, Why I Got Up At 0255 This Morning:

Feline-Induced Early-Morning Insomnia.

It's what you get when the Arcturan Mega-Kitten is awake and thinks you should be too.

No, closing the bedroom door (with the critter on the outside) was not an option.  I've tried that one.  He does his Sheldon impression: (Knock³ Mrow-mrow?)³ until I open the door.

System considerations

That thermocouple dingus a couple of posts down: it uses the Microchip MCP9600 Thermocouple-to-I²C chip.

Which has some nice features: can be set to work with many different thermocouple types, has programmable alert outputs, and so on.

But...

Yeah, the alerts seem like a nice feature.  But ya gotta have an MCU to set them in the first place, and how hard is it to have the MCU wake up periodically, read the temperature, and check the limits?

And then there are the fiddly details.  Like, the T+ and T- pins are backwards with regard to a standard thermocouple socket, so if you're designing in a socket, it's necessary to cross the (very sensitive, low-level analog) traces.  (I've also seen this on some chips with USB interfaces: gotta swap the traces between the chip and a normal mini-B or micro-B connector.  Which I guess is OK at 12 Mbps, but if any 480 Mbps parts have this issue, it could be a Big Freaking Deal.)

And: no open-sensor detection on chip.  Putting a 1M pullup to +3.3V on T+, and a 1M pulldown on T-, gets an off-scale error if the thermocouple is unplugged... but it also introduces an error of a few degrees C.  Maybe 10M would work better?

There also seem to be some interesting issues with the I²C bus implementation, but I'm not sure it's worth my time to verify those.

So... the alternative is the older Maxim MAX31855K, in a boring little SOIC, with an SPI interface.  Not a lot of features, but the sensor connections are right-way-round, and it has built-in fault detection.  Only does K-type thermocouples (other types call for different part numbers), but since my board has a K-type socket on it, and K is what the world uses, that's not a big deal.  Oh, and it "only" has 0.25°C resolution... which is considerably finer than the actual accuracy anyway.

I have one of Spark Fun's MAX31855K boards, and had a Kinetis FRDM board talking to it back in November, so I know it can be done (and I still have the code).  It does one thing, properly.  Win?

Main downside is that the MAX31855K is a little more expensive than the MCP9600, in any given quantity.  But, for having a more suitable feature set, and generally being better suited to the task at hand?

I think the production board will use the Maxim part instead of Microchip.  Most of the firmware work doesn't relate to the actual sensor interface, anyway: it's USB and SCPI and general core stuff, with the sensor part being nearly trivial.  So the board on my workbench continues to be a testbed and a training exercise, and sometime soon I'll take the hour or so to change the design to use the other chip.

Afterthought: Before I do that redesign, I should attend to another task.  The next iteration is meant to include a nice button-conditioning device to differentiate among (a) brief button presses for the firmware to worry about, (b) reset, and (c) reset into bootloader mode.  I need to program an ATtiny9 to perform this function, and ensure that I can actually flash the buggers (in the great bulky SOT23-6 package, because I couldn't find the DFN programming socket at any reasonable price).

Update: This project returns to the back burner for now; a long-awaited pile of paying work came in just after notional closing time on Friday, and that gets priority.

DesignCon report (-ish)

Well, not quite having finished even the first pass through the show floor on Wednesday, I returned Thursday to catch the last ¾ of a row and do the second pass (in reverse order, scanning for things I might have missed the first time).

I mentioned that the overflow - actually, the primary these past many years - parking garage had been partially closed on Wednesday?  Well, on Thursday it was completely closed.  Not that there were any signs to that effect.  There was someone standing by the entrance redirecting arriving would-be parkers to the stadium/amusement-park lot across the street from the convention center.

If their traffic-management planning is that bad on a normal day with a medium-sized event at the convention center, I don't want to be on the same planet on game weekend.

And there's a mess of big construction going on at the stadium.  Is the NFL paying for it, or do the taxpayers pick up the tab, under the new definition of "public works" that includes things built for the exclusive benefit of certain for-profit businesses?

Anyway: the show itself, while much diminished from a few years ago, was still interesting; I gathered a fair stack of paper (mainly relating to contract manufacturing services, which is something I'll be needing to know about Real Soon Now), and did quite a bit of social networking (the real kind, not the Facebook nonsense).  Picked up some ideas, and possible business leads, and got information on potential suppliers and such.

I also crossed paths with two former co-workers (one still local, one now working in Texas), and did a little catching up, to be continued via e-mail.

Trendy this year: cloudy CAD tools.  Also, free-in-some-situations versions of the old expensive tools.  I may check some of these out, in my copious free time, though for my purposes EAGLE has been pretty good.

Oh, yes: I went around hassling connector vendors about my USB type C connector issue.  There are all these companies coming out with connectors for this wondrous new standard, but every dang one of them has a PCB footprint that requires tiny little plated slots for the stamped tabs of the shield: nothing's fully-SMT nor designed for round holes.

This is fine for volume production, but in prototyping the plated slots mean the difference between a $200 (or even $20) budget prototype fab run and a $5000 full-service fab run with major headaches in design, communication, and setup*.  Why are there no prototype-friendly type C connectors for the shoestring-budget startup?

Well, the Würth booth included the guy who fields this sort of question, and it turns out that I'm far from the first to ask it.  Seems the connector footprints have been Standardized by the USB Central Committee.  As in, the latest iteration of the idiots who figured 65534 vendor IDs ought to be enough for everyone.  Being totally isolated from the world of garage startups, they didn't even consider prototyping costs.

Yeah, having the footprints standardized is a fine idea, and it ought to mean that each configuration will be multi-sourced.  But excluding anything not thought of by the Committee leaves some of us out in the cold.

Oh, well.  Should I decide I actually want to use those things, I guess I can fire up the Bridgeport and make a jig for bending the tabs out to create an SMT part.  Maybe.  Or just bend the tabs with pliers.  Meanwhile, I'll be using old familiar mini-B.

On my other connector issue - basically, I want something resembling SMA-for-twinax, or otherwise a good way to get single 100Ω shielded twisted pairs into and out of a box - I found one maybe-sort-of-plausible solution, and a lot of options for plugging in a pair of 50Ω coax cables, which seems to be The Way It's Done.

* Regular drilled round holes, plated or otherwise, are handled in the CAM files.  Slots may be communicated in the Gerber file for the board outline, but this requires out-of-band communication to establish that that's what's going on.  Plated slots need to be identified on the fab drawing, which means someone needs to read the fab drawing.  Also, plated slots require at least one additional manufacturing step, so the board design can't be run as part of a budget panel.  Little tiny slots require use of a little tiny cutter, which again wouldn't normally be used in making a budget panel.  All of this adds up to: I Hate Plated Slots, and will go out of my way to find a connector that doesn't require them.

On my workbench

Yes, it has a K-type thermocouple on one end and USB on the other.  It doesn't speak SCPI yet, but I just got the firmware to the point where at least the USB stack is behaving itself, it doesn't have a blatant resource leak, and the get-the-work-done task can read from one ring buffer and write to another without having to worry about the details of USB... and it can read the temperature of the thermocouple, too.  Migrating my SCPI parser should now be easy.

I had to unplug the USB cable for the photo, because the LEDs were dazzling the camera.  I should probably increase the value of the series resistors a bit.

This particular gadget is basically a training exercise (though I may put it to work in my own lab, should I need to feed temperature measurements into a computer), but the firmware should be useful for Real Products.

DesignCon non-report

Spent yesterday afternoon there.  It's a bit small - maybe half the size it was in its heyday - but I still didn't make it all the way through the expo hall in the 5 hours or so between my arrival and closing time, because I kept stopping to talk to people.  Funny how that works.  Have to go back for a while this afternoon.

Parking is even more FUBAR than usual.  The usual entrance to the overflow garage was closed, so once I reached that entrance and saw the closed sign I had to make another U-turn and go back to the other entrance.  Then it turned out that most (all?) of the parking on the first floor had been blocked off.  Apparently something to do with owls?  The big sort of dish-shaped structure across the street is being prepared as an owl nest or something?  Oh, Superb Owls.  Those things.

Yes, not only did our esteemed rulers decide to spend our money to put a stadium in an area that already had traffic and parking problems, they've now decided that holding the Superbowl there will increase their prestige even further.  Hurrah!

Update: Oh, yes.  While the game will apparently be played in Santa Clara, it seems that the major social events will be in San Francisco.  So we can expect 101 to be a parking lot all weekend, for the entire length of the Peninsula?

Someone just suggested listing my house on Airbnb for that week, and making an easy few kilobucks.  Which I might consider, were the house not a cat-filled disaster area at the moment.

Somewhat related to the previous post

I've been looking at the Cryptonomicon,
Just been glancing at a page or two

OK, so I'm re-reading it.  It's been a few years.  I felt the need to refresh my memory regarding something pertaining to Qwghlm, the nature of the inquiry now having entirely slipped my mind, but it had nothing to do with illuminatory contrivances.

Anyway, I'm not up to that bit yet, but while waiting at the tire shop this morning, I noticed the characterization of Charlene's crowd (starting on page 80 of the edition I have), and the note that nobody in the real world took them seriously.

This was written in the late 1990s.

Fast forward to now; Charlene is in a position of local power, and the denizens of the Halls of Power take her crowd altogether seriously... or at least pretend to.  And the Taking Them Seriously business has infected the corporate world - consult your HR department - so even those who live in the real world and do real work are subject to the insanity.

See what happens when you humor them, or ignore them and hope they'll go away?