First, question whether you need random noise or simply
a stimulus, to gauge response. An impulse might do as well
as anything.
Second, if you can't get a random function working within
your simulator, you can make a pwl table source and gen
up a table full of pseudorandom values. Or you can mix a
group of uncorrelated voltage sources in series, or current
in parallel, and obtain a pretty chaotic, although not strictly
random, waveform without much complexity.
Third, in-application you will probably have a nonrandom
noise environment, and understanding its true nature and
modeling it may show you things that a truly random one
does not (esp. if noise is synchronous to circuit activity,
phase offset can really modulate the behavior and so on).