Journal98 oct
From Simreal
October 1998
Wed, 7 Oct 98
I spent the week of September 21 in Georgia... but the Saturday before I was able to weld all six of the fore-legs, and the Sunday after I was able to do some finishing on those, plus tack-weld twelve panels for the main legs. Then I spent the next week in a mental vapor-lock... though I managed to finish the layout for controller's I/O board.
The weekend of the 3rd, I was forced to remain out of the shop due to my wife's birthday.
Because I've been doing so much (and my day job has picked up with an intense project we must complete by the end of October) I have resigned from my position as Information Coordinator for the SORC. I intend to stay in touch and be a part of the SORC but with my other obligations overflowing all around me, it wasn't fair to hold a post that I wouldn't be able to fulfill.
I called my machinist on Tuesday... he's been working on my stuff "in the background" so... anyone's guess when it will be ready. He has been re-drawing the parts for running on their CNC machine. The first part (and one of the more complex ones) is almost done... this is the cylinder bracket, so that will be nice to get. It doesn't sound like there is any chance of getting all of the parts this month, but I may get enough so I'm not stalled in construction. And I don't have the money right now to hire-out more machining work (it would cost upwards of thousands).
The Boris controller boards are nearly layed out. The I/O board was a bear, but the MCU board was fairly simple. I should be able to send away for the boards and parts soon.
There are two things I am doing differently on these controller boards. First, I am using surface mount components which gives me more silicon power in less space... but soldering these by hand should be interesting. Gotta get me a teeny-tiny little soldering tip, and some teeny-tiny solder, and probably some flux. I'll let you know when I do it.
The second difference is, this controller is an Atmel with a big chunk of external RAM and all of the peripherals are memory-mapped using a parallel bus. Routing 3 SPI wires is a heck of a lot easier than routing 16 address and 8 data wires. Lots. Way. Those hardware guys sure earn their money. Some of the nice things about this design are: the Atmel allows me to program the system in-circuit (I've added a jack for this), plus I've allowed for an open architecture (eg: the bus is extended to a header) so I can add any number of memory-mapped add-on boards down the road. Hard, but fun.
Next up: Electronics orders and complete welding the leg frameworks.
Finally, you may have noticed some changes at the front page. It is undergoing extensive revisions of the various pages that contain pictures and information. When they re-appear next week, all of the pictures will be nicely organized, and I will have posted all of the CAD drawings, schematics, and other information I have to the web. Things had been falling greviously out of date, and it was hard to find stuff. Heck, it would be fun to index the Journal pages too... but I don't think I'll go that far.
Tue, 13 Oct 98
I finished the re-building of the picture and plans pages... except for the software section. That will have to wait a bit, but when its done it should make more sense and be easier to use.
I'm currently building up enough nerve to send for the first prototype boards and parts for my brains... it will set me back a couple of hundred dollars, but then heck -- that's what the fittings for the pneumatics cost!
I finished the major leg sections (for welding the framework at least)... lots more to do before I fit them together, not the least of which is getting the fittings and brackets! Some of these I'll just make myself next weekend -- others will be coming from Shane Hazelwood at Davis Technologies in California. Eventually.
See the pictures pages to get an idea of the size and scope of this monster robot... all six legs and their pneumatic components laid out on the floor of my apartment.
Thu, 22 Oct 98
I received the prototype circuit boards! They look very nice, though the prospect of soldering onto those tiny little pads is a bit intimidating.
I have all of the parts except for the connectors -- discretes, surface-mount chips, that sort of stuff. I did a run to the local electronics store to get some odds and ends. Finding appropriate connectors has been terribly difficult! I made the mistake of designing the board around the Molex web site and their connector specifications. This is great, so long as I buy the little buggers in 1,000 quantities; otherwise, it turns out that most of the parts I spec'ed aren't held in stock by my various suppliers. Oops. So I zipped through the Newark catalog and found analogues, replacements, and hacks to fill-in those parts. With any luck they will arrive tomorrow, and I can put together the board this weekend.
Surface mount chips are amazing... and small.
My plan for soldering the surface-mount devices is to first super-glue them into place (I heard about this somewhere). Then, using a pointy soldering tip and some very fine solder, to solder each pad very carefully under a magnifying lamp. If I dug up some paste (or possible liquid?) flux, in theory, I wouldn't need to add solder -- just "re-flow" the tinning on the board's pads up onto the chip. I may try that too...
I'm still making things for Halloween -- decorations for our party. I was going to make some "animated" props, but I don't think they will get completed in time. Too much going on this year. After Boris is done, I am going to spend most of next year putting together equipment and props for Halloween '99. A guy's gotta have a project, after all!
I didn't go to the shop last weekend to make the final leg parts -- really no point yet, since I can't assemble them until I get the machined pieces out of California. I'll probably call Shane again next week and see what is up. If no real progress then, I'll start making my own parts in November. So much for my October mechanical assembly deadline! There is still some small hope of getting Boris operational by the new year, but I am going to be moving to the new house sometime this year too...
Wed, 28 Oct 98
Well, soldered some SMD last week. First try, I put the 8515 MCU on a dot of superglue... push, prod, fiddle, and eventually it lines up on the pads on all four sides. <whew>.
Lessee, let's try the soldering-iron-and-solder bit (didn't find any flux that made me happy, really). Heat up a pad for a few seconds, touch it with the solder. Hmm. Not bad. Do it another 43 times. Hmm. Ugly. With a capital UG. Can you say "solder bridge"? I thought so! Neither my soldering tip nor my solder is fine enough for this work.
So I try a zillion things to remove the solder, until I find some fine stranded wire that I smear with flux (cheap stuff from Radio Shack, works better than I thought) and use as a solder wick. <whew>. Looks better.
Not being at all satisfied with that soldering job, I next dropped some simple logic chips down (74hc00, 74hc02) onto their super-glue spots. Easier to adjust on the pads. Then, I paint some flux down the pads and pins. With the iron, I touch the junction of pad and pin and watch as the solder crawls onto the pin. Cool! A quick look with the 10x maginifier (gotta have a jeweler's loupe doing SMD)... some look dry, so I flux and heat again. Hmmm.... not bad. I worry a bit about dry pins, but I don't really want to attack it with my solder. Talking (well, via e-mail) with Randy Eubanks, he says he uses water-soluble flux and smears them on the pins before placing it down. That would probably flow better, so I will try some chips with that technique next. Gotta get some water soluble flux to simplify my cleanup, too... the Shack isn't that high-tech.
Plugged my board into the programmer via. a custom cable... powered them up... software... failed. Hmmm.... the ~RESET line is low... hmmm... switch in wrong. Okay.. fixed.
Plugged the board back into the programmer... power up... software... erase... program... verify. Verified OK! OH MY GOD! It WORKED! <ahem>... well, hmmm... but my little program isn't lighting my LED indicator. That's okay -- getting it to program & verify is a good start. From here on out, it's just one tiny little baby-step of testing after another. Soon, the entire board will be working!
Aha! Found the problem... I was setting Port B and not Port D (a bit of "programmer's dyslexia" I guess). But now the dang board won't program... jeez. Ah. I'm using my crummy RJ45 cable that I crimped myself -- and damaged the snap-in doo-hicky. When I hold it in the socket by hand... program! Run... works! Okay! I've programmed and successfully run a test on the Boris Atmel chip. Damn. That's great. Now to start testing out to other chips and systems... so enough of this blow-by-blow; I've got work to do!
<some time passes> Time for my weekly post. The board isn't really operational yet, but I've just been poking at it in my "spare" time. I've been tired this week, so it's slow going. My next test is to get the DUART to set the indicator lights to a particular pattern... it isn't working yet, so now I step through the wires on the board one at a time and verify they are working right. That's a bit tedious, but so far things look fairly good; even though some of the low address lines seem very noisy, the address decoder is making clean work from them. By next week, I'll have a lot more news about what does and does not work on the Boris Brain Board!
