Silicon Valley Redneck

And so I come skidding to the end of the year.

Yes, it's the last day of the year.

Back at the old old job, some years we'd work late on New Year's Eve, frantically finishing up a project, boxing up the deliverables, and making a mad dash to the UPS office on University Avenue before it closed.

I'd kind of expected to do that this year, only without the tangible deliverables, shipping, and so on.

And, given the realities of the current work, I figured I might fudge the end of the day a little.  Like, work through December 35th.  (Well, the 34th; there's a family gathering on the 35th.)

But, no.  I'm not getting much done today, and, right now, the difference between billing a few extra hours in this month and just letting them be in next month isn't really that important.

Cash flow this year has been appalling: a big chunk of the first part of the year went to no-revenue efforts to sort through the remains of a defunct company; I took that month off in summer; a fixed-price project took way too much time; much of what ought to have been work time was consumed by family crises, low-grade illness, and whatnot; and then there were vacation expenses, including an unplanned car rental, and there was the new water heater, and the very expensive repairs to the Prius.

Still, there's enough cash in the bank to keep me out of trouble for a couple of months, and enough work in the pipeline to keep me gainfully occupied for a while yet.  So, unlike so many people, I'm not finishing the year in dire straits.

And maybe, since Huckleberry and Southmoon are so disappointed they didn't get a big ugly fountain for Christmas, I'll splurge on a little fountain for them for Twelfth Night.  And I did splurge on a new (cheap) ushanka; not only is the old one a bit small for me (fine for costume wear, not so good for pulling down over the ears), but I don't know where it ended up after I took the winter stuff out of the trunk in preparation for the summer trip.  New one promises to be a slightly loose fit.

But: once again, no lavish year-end CAPEX this year.  I'd thought I'd at least upgrade EAGLE to version 7, which has some useful new features, but that can wait until I have a project for which the new features would actually be useful.  (With the new XML file formats, being totally in sync with clients and subcontractors has become rather less important than it used to be.)

As for resolutions for next year... pretty much the usual.  Bugging out of Silicon Valley has been on the to-do-soon list since about fall of 2012, but I didn't even manage to get away for a preliminary scouting mission until this past summer.  Urgency is creeping up, what with alarming (indeed, downright Wagnerian) economic trends, but meanwhile my mother's health is not so good, and I don't think I'll be persuading my parents to tag along even as far as a nice suburb of Nashville, never mind the remote countryside, so I may be tied here by family obligations.  Fingers crossed for the continuation of the local Potemkin economy....

Ah, well.  Time to log the couple of billable hours I put in today (trying to track down timing issues in an FPGA; seems most of the really-worrisome ones involve interactions between the 100 MHz and 300 MHz clock domains, despite all the clocks being derived from a common reference, and fiddling with the phase settings yields results that don't appear to make sense and that certainly don't improve the margins any), and spend a short evening winding down from the day.  Then an early bedtime, after snorting some Neosporin and cranking the humidifier up to full blast.

Yes, an early bedtime on New Year's Eve.  Why do you think I blew the noisemaker on Vladivostok time?

Update: Either the neighborhood was uncharacteristically quiet last night, or I managed to sleep right through the fireworks.  Haven't been outside yet to inspect the shiny new local year.

Happy New Year!

Somewhere in the world, anyway.  It may be 0600 on New Year's Eve where I'm sitting, but it's 0000 New Year's Day in (consults timezones) Vladivostok!

(Blows noisemaker, finishes morning grape juice.)  Hurray!  Now I don't have to stay up until midnight local time!

End-of-year post to follow, sometime this evening local time.

Offhand, I'd say someone lacks perspective.

According to a machine translation of this article (via El Reg, unrelatedly), a circus was closed, and the owner charged with fraud, in Italy, because:

Some shaggy dogs dyed in panda patterns were on display as purported dog/panda hybrids.

Wait.  You mean to tell me that a circus owner fibbed to the rubes?  And induced members of the public to part with their hard-earned cash with things that were not as they seemed?

Isn't that what circuses do?

Now I'm thinking of one of the early Discworld books, in which the then-Patrician of Ankh-Morpork (if memory serves) had decreed severe penalties for writing things which were not exactly and literally true, with dire consequences for metaphors and such.  Also, a perennial favorite Dickens line springs to mind.

Interest rate WTF

Just seen in somebody else's sidebar:

Wait... how does a 15-year, fixed-rate loan have a lower interest rate than a 5-year adjustable?

And what kind of monetary policy leads to this?

If Agatha Christie and Ian Fleming had a love-child...

Joy discovered Miss Fisher's Murder Mysteries on Netflix.

All rather entertaining, so long as one keeps one's mind firmly turned off.  Sort of a gritty, action-packed-with-extra-subplots take on the traditional English teatime mystery.

Start thinking, and, well:

Sometimes the culprit (or one of them, anyway) is instantly obvious based on dramatic convention and blatant clues.

Many plot devices make no sense at all.  That cornet-mute dart gun is absurd (as in: would not work, for multiple reasons), and, in any case, what happened to the dart?  And: given two lads whose only resemblance is the striped coats they both wear, how is it that one can impersonate the other in jail without anyone noticing that his hair isn't even the right color?

Then there are failures of elimination.  If a hefty murder victim's body is found strung up high above ground level, with no lifting equipment in evidence, and your prime suspects are a skinny dweeb and a little girl, there's something seriously wrong - and no one thinks to mention this.

And, reflecting on the social aspects: Phryne Fisher is the very epitome of a skybox socialist, a wealthy member of the aristocracy who user her wealth and power to indulge her every whim without risk of legal consequences, while conveniently espousing every political cause popular with Progressive scriptwriters of 85 years hence.

Oh, well.  At least she's in Melbourne, so there's enough excess population to have a person or two murdered in her presence* every week.  Miss Marple's murder a week would have wiped out her little village in no time.

* Morse's Law: There's always a 50:50 chance that the man who found the body did the deed.

Urgh.

Woke up a bit early after rolling an 18 for Morning Headache.

It's a combination sinus headache and crunchy neck.  I'm guessing it starts as a sinus headache, and then for whatever reason I sept in a bad position.

It's cold outside (by local standards; 44°F), and I left the hallway fan positioned to blow upward and toward the living room, to redistribute the hot air that pools on the hallway ceiling above the wall furnace.  Thus, the furnace has been running nonstop.

The carbon-monoxide detector isn't registering anything.

But...

I also left the vaporizer running all night, and the bedroom window has nary a trace of condensation.  So, indoor humidity must be really low (given how cold the window is).

So, perfectly normal sinus headache brought on by some combination of excessively low humidity and the combustion products of household dust (including cat hair) that finds its way into the furnace.

The vaporizer is about worn out; maybe it's time for a new one, that doesn't have such scale-encrusted and eroded (so cleaning the scale off doesn't work so well) electrodes.

Update: Evening now.  Got a new vaporizer (one of the cool-mist ones, with the removable tank, because that's what was on close-out; Target didn't seem to have the old-fashioned kind).  Never did lose the headache.  As outdoors turns cold again, and the furnace resumes continuous operation, my sinuses return to their horrible state.  Also, I seem to have a slight fever: not good at all, and it suggests that the dry indoor air isn't the whole problem.

Here's a concept for ya

"Fractional reserve blood banking."

Update: That particular thought was brought on by something in the morning funnies. Here we go: this morning's F-.  But...

Now I remember another concept, from yesterday, following from the observation that ATMs only dispense twenties.

Print-on-demand money!

Well... banks can create dollars, right?  So why can't they commit those dollars to paper?

Don't think too hard about it.

Chaotic systems have nonlinear regions

Yesterday's plans got scrambled, for one reason and another, and I ended up driving all the way to Sacramento, starting from Palo Alto around 1130 on Boxing Day.

Not long before that, I'd checked the traffic map, and everything was green everywhere.

By a little after noon, though, there were long stretches of parking lot on the freeways.

And, between observed patterns and checking the traffic map while pulled over for message-checking, safety naps, and whatnot: the transitions between heavy congestion and smooth flow appeared entirely absurd, as there were no major feeders at the transition points.  And yet there were stable patterns; basically all afternoon, eastbound traffic on the 680/80 path was congested from Benicia to Vacaville, OK within the Vacaville city limits, congested again between Vacaville and Davis, and then clear again from Davis onward.

The only way that makes any sense at all is if the difference between normal flow and congestion is only a few cars: that traffic running right at capacity moves at the speed limit; very slightly beyond capacity, noticeably below the speed limit; and, with the addition of just a few more cars, it slows to the point that nobody's getting anywhere and things just pile up.

Which, on reflection, makes a heck of a lot of sense.  Congestion is a positive-feedback function, after all, so once it starts it'll get worse in a hurry.

So next time your drive is horrible for no obvious reason, and traffic gets better and worse with just a few cars exiting or entering... it's not really random.  The road is just right at its capacity, give or take a few cars.

Happy Boxing Day!

For those whose family gatherings dribble on for many days after Christmas, this is a grand opportunity to hit the sales in search of a bargain on the perfect present.

Surely there's a store having a 2-for-1 sale on hippopotameese.

Spam respite

As workplaces shut down for the inter-holiday break, the torrent of spam slows to a trickle.

Or, actually: there's still rather a lot of spam coming in, but essentially all of it is getting bounced by SpamAssassin.

This points to there being two flavors of spam, the easily-filtered and the harder-to-filter.

And the hard-to-filter stuff is being sent by way of work computers on company networks.

Just watch: the volume of spam getting past the filters will pick up on Monday, then drop off again, then return to normal by January 6th.

But, wait: shouldn't the spam have dropped off last week, with all the computers taking Robonnukah off?

Update: Here it is Monday the 29th, and spam volume is picking up again, as I have foretold.

Well, that was a waste.

That prototype board I assembled yesterday evening?

Looks like I skimmed through OSH Park's "This is how each layer of your board will come out; is this really what you want?" page without really looking.

And, it seems, the version of the board I'd uploaded was the version with only the critical manual routes done, and not yet autorouted.

Thus, mighty few of the intended connections are actually present.

Grumble grumble... and neither the boards nor the key components were all that cheap.

Looking back at OSH Park's idea of what I ordered, yeah, that's exactly what they delivered, and it's exactly what the file I uploaded calls for.

Oh, well.  It tests the component footprints just fine!  (And there's one that could use some adjusting.)  And, with a few patch wires, I think I can test at least some parts of the design.  As for testing the whole thing, well, running patch wires to a DFN is a tad challenging, so I don't think the DAC gets wired up on this iteration.

Update: Not only does this point up the importance of being fully awake when sending designs out to fab, it's also a reminder that installing the components for the power section first, and then building up the next section after power is checked out, and so on, is a good approach; it reduces stress on both components and engineer during bring-up, and saves the cost of components installed on a board that's just not gonna work.

Also, I find that, in all the juggling that went on with this design (as I shuffled things around and deleted a bunch of stuff to make for a cheaper test design), I somehow ended up with the transformer primary wired backwards.  Oops!  Well, that's why I start out feeding it from a current-limited bench supply; it gives me time to find these things before the regulator self-destructs.  So, that's two cuts of existing traces, and two more patch wires.

Really, the projects I do for clients never turn into this sort of clusterfail.  But, tinkering with the design in odd bits of time, and making bunches of changes, then not having any sort of review process: most of the time my little personal-time gadgets work, despite all that.

Update 2: When I send the board back out for re-fab, I'll probably change the ADC and DAC to fast, 12-bit parts: cheaper, and suitable for using an MCU to do real-time control-loop math.  (Actually, the Kinetis K64 is quite capable of doing the control loop with 16-bit measurement and 12-bit output, and it has handy host-computer interfaces; alas, it lacks galvanic isolation between the analog I/O and the host ports, so I can't just use one for both sides of the gadget.  I might just look for something with fast, high-res A/D and D/A, in a small package, and use that to replace the ADC and DAC chips, worrying about fancier firmware later.)

Update 3: Actually, a Kinetis KL25 might be just the thing, or at least Close Enough.  Got a fast 16-bit ADC and a 12-bit DAC, plus the processor, memory, and so on, and comes in around the price of a standalone 16-bit ADC.  Also, I have the eval board already on hand, so it should be easy to play with it, check out the analog I/O and the I²C, get up to speed on the mbed environment, and so on.  A 32- or 48-terminal QFN, according as how I feel the need for a separate DAC reference input, ouught to fit nicely.  Too bad it won't run on 5V, but that's not really a big deal.

Update 4: Yeah, looks like a KL25 (48-terminal package, alas) plus a precision voltage reference ought to get 'er done. The mbed library speaks I²C slave, so probably not much code to be written to support having the whole thing look like an I²C memory device, configured by writing data to suitable addresses.  Gotta remember to include some actual nonvolatile memory for storing calibration data, just in case this ends up as a pluggable module.

Update 5: Been playing a bit with the FRDM-KL25Z board and mbed, and it looks like a plausible approach to controlling the critter.  Probably patch it into my oopsie board, bypassing everything except the power supply and a couple of the op amps, and have a go at implementing the simpler forms of the control loops using the mbed environment (which isn't as aggressively real-time as AGROS, but it is supported on this hardware, and will probably get the job done).

Doing math on analog signals

There are chips for that.

They're expensive, and tend not to come in handy little SMT packages.

Also, the last time I used one, it was very particular about supply voltages.

Then ya gotta worry about all the things that affect accuracy....

Well, it actually occurred to me, last time around, that the particular job at hand could have been done much more handily with an MCU, two channels of ADC, and one channel of DAC, plus scaling to deal with the required I/O levels.  But, the client wanted a clone of the original all-analog module, so that was how it had to be.

This time... that board I'm prototyping implements only a subset of the functionality one would wish in the product version.  For the full-featured version, it needs to do a modest amount of math on analog signals, in real time (bandwidth of maybe 100 Hz, though several KHz would be better).  A quick look at Digi-Key doesn't turn up a suitable analog math chip at anything like a price I'm willing to consider, even without looking at details of the specs.

So, perhaps an MCU.  Actually, an MCU driving the current prototype might be able to keep up, except that this one is designed for precision rather than bandwidth, and the ADC and DAC don't really run fast enough.

Final version, though, maybe I ditch the 16-bit conversions, settle for 12 bits with a high sample rate, and use either an MCU or a small FPGA to implement the various options for control functions.  I should look more closely at application requirements; 12 bits might in fact be plenty for most uses, or perhaps I could play games with range selection, or run the control loop at 12 bits and take measurements more sedately at 16.

(Yeah, it's the core of a piece of test equipment.  The real ones are frightfully expensive.  I happen to need one of modest capabilities, hence the project.  Other people might have similar needs, hence the coyness until I decide whether it's going to become a product or not.)

Earworm

As I was slaving over a hot soldering iron this evening... OK, the soldering iron was turned off; I was putting solder paste and components on a prototype board, then reflowing with a hot-air dingus... there was a tune running through my head.

Went from "vaguely familiar" to "Brahms" to (about the time the solder was starting to melt) "Academic Festival Overture"*.

No idea why.

Anyway, I now pass this earworm along to you.  May it dispel any unwanted trying-too-hard performances of commercial holiday songs.

(May have a report on the prototype eventually.  It's a proof-of-multiple-concepts, all rolled into one gadget that's potentially useful... if all the concepts pan out.)

* Come to think of it, it probably started out as Smetana and morphed into Brahms.

Who's been sleeping in my porridge?

Awwww... Huckleberry's trying to be me.

Afterthought: Those critters are getting altogether too anthropomorphic.  Is it not written, No animal shall sleep in a bed with sheets?

Cat logic

The mouse sits on the mousepad and gets petted all day.

Therefore, the mousepad is a place for getting petted.

On hitting people with bottles

I've mentioned, a time or two, that hitting people with actual glass bottles (as opposed to Hollywood props) is several kinds of Not Funny, and likely to cause serious damage to the people hit.

On the other hand, there's this:

That's when officers shot the unarmed Chadwick in the back of the head with a Taser at point blank range.

"They claimed I drew down with a shampoo bottle and a body wash bottle," said Chadwick.

Uh... guys?

Last I heard, bottled bath products came in plastic bottles.  Being hit with a plastic bottle is, in general, no big deal.  'Cause they're soft, see?

Yes, being hit with the heavy-ish contents of the bottle may result in a bruise.  Maybe worse, in the case of a particularly hard blow in the wrong place.

But, when the party being menaced is a cop in full SWAT protective gear?

Really?

Superintendent: Oh sorry! Right, here we go. You are hereby charged. One, that you did, on or about 1126, conspire to publicize a London Borough in the course of a BBC saga; two, that you were wilfully and persistently a foreigner; three, that you conspired to do things not normally considered illegal; four, that you were caught in possession of an offensive weapon, viz, the big brown table down at the police station.

Judge: The big brown table down at the police station?

Superintendent: It's the best we could find, m'lud

Afterthought: Also, bath-product bottles don't tend to be particularly club-shaped.  Hard to get a good swing with one.

May have to learn Python

The programming language, that is.  Not the TV show.  I learned that back in high school.  Had many of the sketches memorized.

Anyway: What with a notable lack of Ruby bindings for Qt5 (which has a nice set of features lacking in Qt4, some of which would be useful for stuff I want to do), I'm sort of casting about.  One option might be to figure out why qtbindings doesn't build for Qt5 (its build script refuses to so its thing; I assume there's some actual reason for this), and fix it myself.

But...

For much of the kind of stuff I'm doing, the rest of the world seems to have gone to Python the past few years (and I seem to be the only one who's ever used Ruby for this).

I actually have a copy of Programming Python: a first edition, in excellent condition, if there are any collectors out there.  Presumably I should consult something a few years newer.

If memory serves, I bought the book because some program I was using had gone unsupported, and it was written in Python.  I think it was Sketch (later reborn as Skencil, but still not available on the current version of Debian, and I still have some old artwork in that format, in need of translating to Inkscape or some such).

I looked at it,  and tried to grok, but just didn't (at the time) get along with the syntax.

Maybe I should look at it again.

I keep having the urge to create a new programming language.  Did that back in college... was called GSPL, and it compiled to 6809 code.  The compiler was originally in SCUD UCSD Pascal, but then I translated it, and built a native version.  (Actually, there were two native versions; the one used to build the compiler truncated symbols at, IIRC, six characters, because otherwise it would run out of memory on a machine with only 56K of RAM.)  I (or, rather, my publisher) sold something like two or three copies; some time later, I actually met one of the users, who'd been developing control firmware for some heavy machinery.

I actually have two, quite different but possibly reconcilable, visions.  Both have basically C-ish syntax.  One is scripting-oriented, and has a more or less Ruby-like (not C++-like, and don't even get me started on Objective C) object system.  The other has upgraded declarations allowing detailed physical record layouts, such that one might declare, say, the structure of a IPv4 header and than just access the fields as normal structure elements, with alignment and endianness taken care of by the compiler.

Last time I took on something like that, it was a C compiler, with some neat language extensions I happened to want (plus 68000 assembler, plus universal linker) from scratch (well, I did have a C compiler, for another architecture, to compile it with).  It took a good chunk of a summer, and I was a lot younger then, and better able to handle the required amounts of Jolt Cola.  Now... well, there are a couple of open-source back-ends out there (gcc, llvm), which would remove much (not by any means all) of the code-generation effort, so it's a smaller project, but still a bit much to take on for a lark.

Anyway, I ought to take another look at Python, if only because it's in common use (and thus a good skill to add to my collection).  I should also see what can be done with Qt's built-in Javascriptish thing; that might, perhaps, be good enough for what I'm trying to do in the semi-near term.

Cool thermal imaging goodie!

Now from Spark Fun!

Well... not now, exactly.  A couple of hours ago, yes.  Sometime in the future, presumably.  Those sure sold out quickly!

Thing is, I've been aware of this sensor (or something remarkably like it) since Fall 2011; going into early 2012, JCM Consulting Technologies was negotiating, via a third or possibly fourth party, for the contract to come up with a manufacturing test and calibration solution for a FLIR module that I think would have looked a lot like this dingus (sans eval board; just the little square-thing-with-lens).  Specs weren't firm yet, timeframe looked a lot like "deliver it last week and we'll tell you what we need in a couple of months", and we didn't land the deal.

If memory serves, projected volume was in stocking-stuffer territory, with a product (whatever is was to have been) incorporating the module retailing for an enhanced-stocking-stuffer price, maybe around $100.

Well, I never did see that product, but now the imager module is (or was, and may be again) available on a little eval board for $350, which is quite a bit more than I'd hoped for, but just having it out there is kind of neat.

No, I didn't snap one up this morning; it wasn't in my budget (though a thermal imager, even a low-res one, is firmly on my tools-for-real-work wish list).  Maybe next year.  And I did have some ideas for products to design around it, so maybe I need one for work and one for tinkering with.

Update: I find someone selling the modules at $300 a pop, and someone else organizing a group buy at a bit over $200.  Also, FLIR now has a complete handheld camera built around this puppy, for $500 from various retailers.

For serious work use, if the need comes up again, I think I'd spring for a somewhat pricier model with better resolution.  But, for funsies, the little low-res module would be kind of neat (and isn't in this year's funsies budget).

Financial news

Monday: STOCKS PLUMMET ON FALLING OIL.

Thursday: STOCKS SOAR ON FALLING OIL.

Makes sense to me.

Speaking of wooden rubles

The crude-oil cartel has been boosting production and cutting prices in an effort to discourage North American crude development.

The rumor mill suggests that Russia is behind this.

In any case, it's a fine example of predatory pricing, and it's been generating headlines.  You've seen them.  Oil prices fall, there's no point in investing in domestic extraction, nor transportation of Canadian crude, all that stuff.  Also, cheap crude will totally destroy the economy (I'm still not clear on how that's supposed to work).

But... isn't there a YouTube vid of an interview with Milton Friedman, in which he explains that predatory pricing doesn't actually work?  I can't find it at the moment, but there's this from Thomas J. DiLorenzo at the Cato Institute.

And what evidence is there that it doesn't work?  Why, after only a few months of selling below cost, Russia is having a grand economic crisis, with the ruble in free-fall.

Maybe there's something to this Austrian economics stuff after all.