Date: Wed, 23 Nov 2005 00:44:50 -0600 (CST)
Subject: motor control board update
X-UID: 107
Content-Type: IMAGE/JPEG; NAME="img2412.jpg"
Content-Type: IMAGE/JPEG; NAME="img2414.jpg"
Content-Type: IMAGE/JPEG; NAME="img2415.jpg"

The circuits for controlling the drive motors is ok.

The 510 ohm 1/4 watt emitter resistors are replaced with 1/2 watt 470 
(really 460 ohm) resistors from Radio Shack. To review, a 4N35 
optoisolator and NPN transistor comprise a Darlington in emitter follower 
configuration. This means that there is no voltage gain, just current. The 
reason is that significant current is required to drive a MOSFET gate 
quickly.

The power MOSFET gates have enough capacitance to have a significant RC 
constant. So the lower resistor increases the switching speed of the 
MOSFET. There is a practical lower bound to how small this resistor may 
be. At a gate voltage 12 volts and 100% duty cycle, a resistor of 2 * 12 * 
12 = 288 ohms dissipates 1/2 watt. The next larger common size are one 
watt resistors which are much larger. The other alternative is to use 
multiple resistors in parallel to distribute the current and hence power. 
I didn't do that because every additional component adds significant work 
to a wirewrapped design. The cost of labor, testing and potential broken 
connections drops with every component that can be eliminated.

Note that a production design would use an IC with charge pumps to drive 
the MOSFET gate. This was my original design using HIP4081s. What I'm 
doing now is simple and primitive. Power consumption in my current design 
is significantly higher. However, this design is still reliable as there 
is complete optical isolation of the logic and power sides. Also, all 
components are kept well below maximum ratings by design.

In the photo of the testing setup, the blue trace is the TTL logic level 
pulse width modulated signal from the microcontroller. The tall yellow 
trace is the voltage across the Mabuchi motor. Note how the yellow pulses 
are wider. There is charge stored in both the power MOSFET gate as well as 
the optoisolator phototransistor base. Both transistors remain on until 
that charge has drained away to the emitter/source. This causes the pulse 
width to stretch. It's not important that the input and output duty cycles 
match for this application.

Not all narrow hole wire wrap sockets are the same. The diameter of the 
copper cup in the pin holes varies somewhat. The ones from DigiKey have 
slightly larger holes than the ones found in a cardboard bin at Altex in 
Carrollton. So I used a Dremel 1/32 engraving cutter to bore out the 
copper cups wide enough for the metal leads of the zener diodes and 1/2 
watt resistors to fit. A hemostat clamped on the leads near the socket and 
downwards pressure forced the components into the sockets.

Next up is the relay coil control part of the board. After that comes the 
cooling fan and headlight control. I decided that the electronics 
enclosure cooling fan should be under computer control. Turning it off in 
cold weather or when the robot isn't doing much could save power. I also 
decided to add a headlight under computer control. It would be nice to 
connect the two lasers to the same control circuit. But the lasers have 
bubble membrane soft on/off switches and I don't really want to cut into 
those. I've had bad experiences hacking laptop plastic ribbon cables so 
fear the same thing might happen. I'll probably leave the lasers alone and 
go back later to figure them out.