The complete circuit therefore consists of a differential amplifier, and a power buffer that must be able to drive the motor in both directions.

During the design phase I had a number of considerations for the circuit:

The Pitch and Roll drive circuit is shown below. It consists of a differential input circuit that provides the error signal,  a sawtooth generator that determines the switching frequency, 4 comparators that make the PWM (Pulse Width Modulation) signals, 2 driver IC's that make the PWM signals suitable for driving the MOSFETS. 

Each axis motor is driven by a half-bridge configuration with dual (4x12V 7Ah battery) supply: Two hefty N-Mosfets IRFP250 that are driven by a low-side high-side driver IC IR2110 
This IC is capable of driving the bottom Mosfet and the level-shifted top Mosfet, which gate drive signal rides on the center switching waveform. Two separate inputs are available for controlling each Mosfet. 

The high-side Mosfet drive has it's own floating supply. You could make use of the bootstrap trick as shown in the IC spec, but I decided for a separate floating supply via a small transformer, as I had read somewhere that the bootstrap application could have some problems, especially during start-up. With the solution shown above, the Mosfet drive always has correct gate drive swing, regardless the switching conditions.

The IRFP250 Mosfets are selected because they have rather fast inner body diodes, that carry part of the inductive current.

The driving signals for the IC are made such that there is some dead-time between the switch-off of one Mosfet and switch-on of the other. Without sufficient dead time, the two Mosfets could conduct simultaneously, shorting out the supply, thereby killing both Fets.

The working of the high and low side Mosfet drive is shown above:
The drive and position feedback are added together. The output of the opamp drives the inputs of two comparators, where the resistor network takes care that there is a small offset between the two comparators inputs. The other inputs of the comparators are driven by a triangle waveform. (actually it's a sawtooth but it does not make a difference)
The resulting outputs of the comparators is shown at the right. Moving the DC voltages up will result in longer on-times of the low side Mosfet and shorter On times for the Hi-side Mosfet  and visa versa. Care must be taken that the comparator voltages stay within the triangle waveform to ensure continuous switching waveforms. The dead time between two transitions will remain constant.

The value of the 220k resistors from the position feedback potmeters together with the 180k feedback resistors determine  total loop-gain, and need to be tweaked to get optimal platform motion response. You can do this by inserting a step pulse on the drive input, and check how the platform reacts. My final value was 220k increased to 820k for reasonable response, without overshoot. In case of oscillation, small capacitors may be needed across the 180k feedback resistors. Note that this tweaking needs to be done with all weight load added to the platform. I used three 20 liter water bottles to simulate my weight. I did not do any frequency analysis on the platform, but it was reasonable fast: A large +30 to -30 degrees movement is accomplished in about 400msec. Small signal frequencies up to about 10 Hz are definitely possible.    

The small signal power supply for the drive electronics is a straightforward +/- 15V supply. The high power supply is using batteries as they can deliver lots of current and most important, also can sink lots of current. When switching on the platform drive power section, it is important that this is done as gently as possible. Since it is a closed loop system, the platform drive will try to correct the platform position immediately after switch-on. To do this in a gentle way, the supply of the half-bridge is done in two steps: After switch-on,  after some delay the half-bridge +/- 24V supply comes up with two 47 Ohm resistors in series via relay A, and after some further delay the second relay (B) comes in, connecting the supply directly to the bridge. The two delay circuits that drive the relays A and B is activated by the platform on-off switch. To be operational, all end-stop switches must be in closed position. As soon as one of the end-stop switches is opened, the supply to the half-bridge is disabled.       

Some pictures of the motion platform drive electronics. The heatsink  does not need to be this big.
Wiring is done via solder, transformer wire and thick wire for the high-current section. The layout of the IR2110 to the Mosfets must be done with care. As always with these kind of switch mode power circuits, keep small signal ground and power ground separate, and keep gate-drive current loops and supply decoupling loops small.

Next