Comparison of Random numbers generator on calculators

[Gil]

Is there a simple way to evaluate / compare "quality" or effectiveness of random numbers generators on two calculators, say HP50G and TI?

Of course, generating random #1, #2... #99, #100, I could build pairs (xi, yi), with x=#1 and y=#2, asf.
Then draw the 50 pairs and see that no pattern is being shown. But it is quite subjective

On the other way, I would tend to say no to that question,
as for 5 trials, for instance, I imagine I could get 5 times the same number. Of course, such special occurrence should be rare, considering the 12 digits of the calculator, about 1 in 10^60 cases.

But, statistically, my temptation would be the following:
Experiment 1
Have random numbers, all integers, from 1 to 100
Proceed to the above 10000 times.
Calculate for HP : (expected-observed)²/... +
Calculate the same for TI.
Compare them and see the "winner" (smallest sum).
Experiment 2: Procede again as in experiment 1 asf.
... Experiment 100... and see how many times HP or TI has smallest acceptable khi2 value.

[cortopar]

Provable randomness has a good outline on this topic.