Since PrISM ises pseudo-random numbers (as the first two letters indicate) there must be a "seed" for the generator. Try to dig that up in the sources and modify it for each of the motors.
Not sure what you are trying to accomplish. Being a psuedo random
generator, PrISM, all 9 can be on or off at the same time, same is true
of x out of 9, seed value is irrelevant from this point of view. "Phase" is
almost irrelevant in the normal sense.
This is true of either density output or psuedo generator output.
If goal is each have a unique state in time of the psuedo sequence of generated values,
then seed value would accomplish this.
With different seed values you will get a different series of pseudo-random numbers. PrISM just decides whether the actual number is larger (LED on) or not (LED off). So in fact it might turn out that all your LEDs are on at the same time. Better would be to stabilize the supply.
Bobs point about a "stiffer" supply prescient I think. Especially in light you
have random processes that can result in worst case maximum load transients.
Some considerations -
1) I assume you are using some form of motor driver. If a simple MOSFET beware
of transient coupling due to Ldi/dt back thru Cgd and injecting spikes into PSOC.
A 2 stage isolated driver offers some isolation. or diode/zener clamps at PSOC inputs.
2) Use a scope, trigger on something below ground, like -1 volt, and look at driver pins.
Same for > Vdd, to see if you can establish where problems lie.
3) Use scope on infinite persistance and look at Vdd rail noise, might be an eye opener.
4) Polymer tantalums have a ESR vs F curve ~ 1 order of magnitude better than traditional
tants. Also use MLC caps for the faster higher freq problems.