building first servo, got questions...

Hi folks, I'm trying to build my first servo, and I've got some questions. I'm building it out of a small brushed DC motor (Pittman 8322 with shaft encoder) and a simple home-made L298-based servo driver (love those free samples). I'm controlling it with EMC feeding it pulse/direction signals over a parallel port.

The motor specs:
reference voltage: 19.1 V DC
no-load current: 0.16 A
Peak/Stall Current: 2.51 A
Continuous Torque: 1.6 oz*in
Stall Torque: 7.4 oz*in
no-load speed: 7847 rpm
encoder: 256 cpr (1024 quadrature edges per revolution)

First of all, does this seem like a reasonable base for a servo motor? Given the limitations of its torque output, of course.

The first step was to read the encoder, that was easy and works fine. The encoder produces 1024 edges per revolution. I'm sampling the A and B lines at 40 KHz, so by my calculations I should be able to reliably keep track of position up to about 1100 or 1200 rpm.

The next step was to spin the motor, and that's where I'm running into questions.

I'm using a driver circuit basically identical to the one published by Jeff Epler. The only power supply I have access to right now is an ATX one scrounged from some dead computer, so I'm driving the motor at only 12 V; it's rated for 19.1 V so that's probably part of my problem. I run a PDM pulse train into the H-bridge enable pin, and direction and not-direction signals into the two H-bridge inputs. The resolution of my pulse generation is about 100 us, so the pulse train has transitions at no more than 10 KHz, well below the 25 KHz recommended commutation frequency of the L298 H-bridge.

At 0% duty cycle the motor is stopped, and at 100% it runs fast, so that's good. However, the mapping from duty-cycle to rpm is not linear. Duty cycles from 0 to about 55% give me increasing whining from the motor but no movement at all. 55 to 100% gives increasing motor speed and torque.

So my main question is: What, if anything, can/should I do about the dead band from 0 to 55% duty cycle? Live with it and tune it out in the PID loop?

What causes this? I guess at those low duty cycles (especially given the low input voltage) the energy being fed to the motor is not enough to overcome inertia or static friction or something.

I tried increasing the pulse-generating period, up around 1 KHz the motor became more responsive but also more "clicky" and rough. I think 10 KHz is about as slow as I want to go.

If I can get this servo moving reliably I'll build two more, and put them on a McWire Mill.