Shown here is how InstaSPIN-BLDC actually works: So what TI has done here is actually replicated the three phase commutated voltage waveforms for a brushless DC motor and for each unpowered phase, the back-EMF signal crossing through zero can be seen. The designer has also added the effects of the inductive commutation whenever a coil turns off. Because there was current flowing in that coil, the current has to keep going somewhere because the coil is inductive and as a result of that, it creates a voltage spike as shown in the waveforms in the diagram. Now, the back-EMF voltage crossing zero during the unpowered phases, instead of actually timing the points at which it crosses zero, what the designer is going to do is integrate the area under the curve after it crosses zero. So just take phase A for example. The first instance where the phase is turn off and the back-EMF signal appears and it has a negative slope on it. The designer immediately starts an integrator which is trying to integrate in the negative direction but it is bounded by zero. So, during the portions of the waveforms that are positive, one cannot integrate because the integrator maximum value is clamped at zero. But as soon as the waveform crosses zero then the integrator starts integrating negative numbers and this is a linear waveform, if that is integrated, one actually gets a quadratic-looking waveform and that is what is used to commutate the motor with. From magnetics classes, what is the integral of voltage? It turns out it is flux. The waveform that is actually being recreated to commutate the machine is flux, and all that has to be done is set a threshold for when the flux gets to a certain value and that is when the machine gets commutated. In other words, counting the flux lines that are being accumulated as the magnets sweep past the coil and once the flux lines have been counted enough, the rotor must be in the correct angle now to commutate the machine. Now, what would happen if the magnitude of the V threshold values were reduced, that the flux waveform is being compared against? That means that the flux waveform is going to hit those threshold values earlier. So, at higher speeds if the designer wants to cause commutation events to occur, all that has to be done is to reduce the magnitude of the threshold values.