So I recently acquired a nice micropump centrifugal pump, with a 6 pole brushless motor. It has 3 hall effect sensors. I recently made a quick program to control the commutation, so next up is speed control. The motor is rated at 36 volts at 3 amps. The transistors
I'm using to control the motr are igbts in an integrated module, so gate drivers and fault detection are covered automatically 
I strongly reccomend them, $25 on eBay, rated for something like 300 volts at 45 amps. Anyway, so the next step for me is speed control. I had 2 ideas. The first was to use pwm to modulated the positive supply line. The next idea was to adjust the length of the motor pulses, as in when the hall effect reaches the next cycle, turn the motor on for a fixed time and let it coast to the next commutation event. Similar to pwm. I've been reading up on motor control theory, nut I would love to hear the opinion of someone with experience.
You probably won't be able to modulate the supply as the module will have decoupling capacitors trying to hold it steady and its protection circuitry will probably expect a steady supply.
Need a bit more about the module to know how to drive it (link?). PWM on one of the active lines is commonly done. There are various ways to do it depending on where you want the current to flow when the normal path is deactivated.
If you want to motor to work at low speeds you can't rely on it coasting to the next commutation event, it might never happen.
Hi, if you can point to data on the modules and motor, we can think more about this.
What's the minimum RPM you want to run? This may decide whether it possible for the PWM to be per-pole-period or if it has to be faster like Mark suggests.
The igbt module for generalization is six seperate igbts w
ith individual drivers. My power supply is a seperate 12v regulated supply. Modulation of the supply would not be a problem, there are no decoupling capacitors to interfere. I do not have the datasheet with me at the moment, I remember the switching speed is limited to about 15 khz though. The more technicial aspects I can solve myself, I just need a point in the right direction to the control theory.