Date: Thu, 10 Nov 2005 23:26:11 -0600 (CST)
Subject: why motor control circuit doesn't work
X-UID: 103

The driver chips have inadequate power. There are two LM2574 switching 
regulators, one for +5 volts and the other for +12 volts. These can supply 
500 milliamps each. This is insufficient. The peak current requirements of 
the power MOSFET gates is between 1 and 2 amps. Compounding this, I used 
30 gauge wire in most places. This limits current to around 600 milliamps.

So what I did was analogous to wiring a house with inadequate current 
capacity. All of the outlets have the right voltage. But as soon as the 
refrigerator, microwave oven, and hair dryer turn on, the circuit breaker 
trips.

For the motor control board, switching the MOSFETs worked initially as 
they were switched into one state (on or off) and then held there. While 
the current was inadequate, I didn't see this as the gates would charge up 
in a few microseconds, allowing the circuit to reach a steady state.

But as soon as the microcontroller switched the MOSFETs 2000 times a 
second, the system could never reach stability. The transistors couldn't 
switch properly. The power problems were also causing the motor control 
ICs to shutdown and behave erratically.

Here's one technical lesson I'm taking out of this. One interpretation of 
electronics is signal processing in which information is represented as 
voltage over time. Digital electronics at low frequency is the extreme 
example of this. All analog effects can be ignored and everything is in 
saturation. Outside of this region, these simplifying abstractions become 
invalid. Power electronics is an example in which large signal analog 
design considerations are necessary.

I read on the Wikipedia that field effect transistors are voltage 
amplifiers while bipolar transistors are current amplifiers. Now I see why 
this is so and why there are so many different device families in 
electronics. It's not a digital universe in which one kind of logic can do 
the work of all others.

The peak current requirements for driving a FET can be high. Other 
components such as digital logic or optical isolators can not source 
enough current to drive the FET directly. So a bipolar transistor between 
them can solve this problem. The computer stuff in the realm of low 
voltage and current controls bipolar transistors which then in turn drive 
the high power FETs.

I'm redesigning the motor control from scratch. This time, I'm not using 
any ICs at all except for five 4N35 optical isolators (two for the drive 
motors, two for the steering motor, and one for the forward/reverse 
shift). There are no power supplies either. I'll just use batteries to get 
the right voltages. I want a robust and simple design that will work and 
not break under hard use. I can live with more maintenance and logistics 
headaches (have to make sure all of those batteries are charged).

One simplifying feature is differential drive for the two wheels forwards 
but not in reverse. The old design had three full H-bridges. It could 
drive all motors in any direction at any speed. This was not necessary. 
The new design will only have one speed in reverse. But that's ok as 
reverse is only rarely used to back out of bad spots. The simplification 
that results is that two MOSFETs and a relay in the new design now do the 
work of eight MOSFETs in the old one.