Date: Mon, 28 Nov 2005 02:22:52 -0600 (CST)
Subject: motor control board done
X-UID: 108
Content-Type: IMAGE/JPEG; name="img2418.jpg"
Content-Type: IMAGE/JPEG; name="img2419.jpg"
Content-Type: IMAGE/JPEG; name="img2420.jpg"
Content-Type: IMAGE/JPEG; name="img2422.jpg"

The board hasn't been rigorously tested. But spot checks of different 
parts are working. I have good confidence it is o.k.

Two accidents destroyed some hardware.

1. I made a mistake and plugged the logic power on the motor control board
    into the 18 volt connector instead of the 5 volt one. This fried the
    microcontroller which was replaced with a spare. However, all other
    components on the board are well within tolerances so are o.k.

     - Lesson: It would be best to have connectors at different voltages
               painted different colors or even better to be of completely
               different type.

2. A piece of metal brushed against some hot terminal block screws and
    shorted 18 volts to ground. This fried the two active GPIO (general
    purpose input output) pins in use at the time on the lower computer
    board. Now these pins are always slightly above ground. However, other
    GPIO pins as well as everything else seems to be o.k. I may have been
    very lucky.

     - Lesson: (part 1) Have to be more careful and respect electricity. A
               careless mistake can cost hundreds of dollars.
               (part 2) Powered circuits and devices are more likely to
               suffer damage in adverse conditions. If a machine expects
               trouble, powering down subsystems is prudent.

I have to see if the boards survived the abuse o.k. I'll only know that 
with some confidence if they can run more complicated software. So in the 
next few days they'll have to be interfaced to the 900 MHz radio and GPS 
receiver and then stressed. Any crashes or resets will be taken as 
failure. In that case, I'll order replacements.

One potential expected problem is the need for buffering the logical 
signals from the computer boards. The GPIO pins are limited to sourcing 3 
milliamps at 3.3 volts. An emitter follower with a NPN transistor was 
necessary for serial communications between the computer and motor control 
boards. I don't know if the RXD/TXN TTL serial port pins will also require 
this.

In hindsight, the Soekris 4801 embedded router boards are not ideal for 
(beginner) robotics applications. This is mostly due to very limited 
power. As a result, the standard PCI slot uses 3 volts signaling. That 
should be ok except many cards are still 5 volts. In practice, some 
hunting is required to find a compatible card (or a card must be modified 
to work - the USB card I'm using required hacking). The GPIO pins can not 
source enough current to drive the logic of a microcontroller. This means 
all pins must be buffered.

I had conceptualized the robot hardware in the following two stacks of 
layers. Note that these stacks are from low level to high level as that's 
the order of fabrication.

STACK ONE
1. vehicle platform
2. batteries, motors and analogue sensors
3. motor control board, the power side
4. motor control board, the logic side
5. single board computers

STACK TWO
1. radio, GPS receiver and WiFi router
2. single board computers

What I'm discovering is that both stacks are missing a layer. The single 
board computers may need some buffering to interface with the digital 
devices above them on the stack. The STK 500 microcontroller board buffers 
all logic to and from the microcontroller it hosts. This is a point I had 
completely missed earlier.