Isolating motor control signals from microcontroller from high voltage/current lines

We need to design an integrated DC motor controller/driver, which will be under microprocessor control. Due to space requirements, either the digital and analog circuitry will be on a single PCB (think of PC104 size), or they will be stacked on top of each other (again, similar to how pc104 boards are stacked). The load requirements are about 10A at 30V. Stall current can go up to 25A.

In order to protect the controller from the motor, we want to isolate the control signals. In fact, even the power supply and the grounds for the digital and analog sides are different. (Note that this means the the ground of the analog side can be at an entirely different voltage than the ground of the digital side)

My questions are: What is a good way to isolate these two domains (optocouplers, maybe?). The signaling rate between the two domains need to be around 1Mhz.

Also, I'm a bit worried about the noise from the motor affecting the digital circuitry, even with proper isolation of the control signal lines. Just being physically close to a motor sometimes even causes problems, let alone being physically connected. I would like to hear about your experiences in building motor controllers/drives to we don't make the same mistakes.

Answer

Optocouplers can be used in many applications but be aware that the switching speeds are limited. Digital isolators (Analog devices, et al.) using magnetic or capacitive coupling are much faster, but slightly more expensive. We have had good luck with all these approaches. Generally, optocouplers (ordinary with external drivers or drivers like Avego HCPL-3120) will do since the switching speeds are rarely over 100kHz on motors. Use gate drivers that have fast enough and powerful enough outputs to keep switching losses under control. For analog feedback consider isolation amplifiers (TI, Analog Devices) LEM's or optoamps (Avego) We will often mace the control circuitry hot and only couple in and out the control information. As far as noise goes, avoid running power through the ground of any control or measurement circuitry. Use ground planed control PCB's with a single connection of their ground to power ground if possible. I have successfully used a small 2 layer SMT (one side grounded) control board on a power system with 83 amps peak at 385 volts and 62.5 kHz and had not even a tiny bit of trouble with noise so far. The SMT is mounted directly on the power devices with short standoffs and gets its control signals through an 8 pin header.