Date: Sat, 20 Aug 2005 01:44:03 -0500 (CDT)
Subject: almost done with electronics
X-UID: 80
Content-Type: IMAGE/JPEG; NAME="img1791.jpg"
Content-Type: IMAGE/JPEG; NAME="img1792.jpg"
Content-Type: IMAGE/JPEG; NAME="img1793.jpg"
Content-Type: IMAGE/JPEG; NAME="img1796.jpg"

The pictures of the electronics board are larger to show detail. It's 
pretty messy. I don't know if it all works. I can't really test it until 
the microcontroller firmware and systems software for the single board 
computers is ready. I'd like to say I can do this in a week. But all of my 
other near term time estimates have been way off (although, the big 
picture project schedule appears to be fairly accurate). However, I've 
much more experience in software than in fabrication (this is my first 
project with moving parts) and electronics (this is my first electronics 
project). And my mindset is pushing very hard now to make progress. If I 
can put in an 80 hour week for a job during crunch time - why can't I do 
this for my robot?

<soapbox>
Some design rules for the future:

1. Modularity with clear interfaces may result in a bulkier design (more
    boards and greater size) but is generally more robust.

2. Isolate different types of electronics if possible. This board mixes
    power electronics with digital and is more complicated as a result.
    This is really a corollary of the first rule.

Note that these rules apply equally well to software as well as hardware.

So why did I design the electronics this way? Because I had no experience 
and saw only circuit diagrams drawn on paper. I had no idea of the 
practical construction, electrical and thermal design issues that are also 
in play. The paper design did not address these other aspects at all. 
They had to be worked out as they arose. The result is shown in the photos.

Here's my deeper question - do these engineering design principles apply 
to mathematics at all? Or is mathematics a pure science? I'm thinking in 
particular of one mathematician who crossed over to an engineering focus. 
He was denied tenure at NC State and went on to Wake Forest and intimate 
work with the USAF on adaptive optics (flexible mirrors). His entire focus 
was fast computation of approximate solutions to optical inverse problems 
(typical applications are spy satellites, optical telescopes, high power 
laser weapons). This inevitably leads to work in which it is not possible 
to have the traditional sort of applied mathematics result - i.e. given 
the specified conditions, convergence is guaranteed to within this bound 
at this rate, etc. I also saw that at Stanford with the device and process 
simulation of semiconductors. Engineers develop wisdom about what works. 
But nothing is known for certain. This seems to be a general theme in 
technology - it is never 100% reliable so never provably correct. This is 
observed in practice with all manner of technology, hardware and software, 
that malfunctions under the right conditions.
</soapbox>

A tip - the oblong Nutella (hazelnut butter - very nice on crackers) 
plastic canisters are just about the perfect size for most smallish wall 
wart power supplies. It fits the 12 volt 500 milliamp supply I use for an 
ethernet switch and the STK500. Also, the Nutella spread washes off the 
plastic pretty easily without noticeable residue. The Nutella canister is 
much better than the Fram oil filter can.

The plastic parts box that is now my STK500 microcontroller programming 
kit had a good deal of the bottom cut away and a thick plastic sheet 
cemented in place. The white frost is due to the sprayed contact cement. 
The black electrical tape is not structural. It is just to cover gaps with 
exposed cement. I didn't want detritus to find its way into these holes 
and stick. That would be unsightly.