Maximum number of data points in a power spectrum transform

[Melmo]

What is the maximum number of data points that can be used in the power spectrum transform.

(NationalInstruments.Analysis.Dsp.Transforms.PowerSpectrum)

The Help file states 2²³(for power 2 quantities), 2²²–1 (for non power2 quantities). Are the non power 2 quantities not just zero padded to a power of 2, which should make the maximum non power 2 quantity 2^23 -1? 

What memory amounts are these values based on, as i know on my system we can use the transform for upto 2^25 samples.

We are using a high rate data capture system so  i would like to make this number as large as possible, before we have to start applying downsampling, however i would prefer to limit this to a quantity that would be resonable for our target machine to handle.

The target machine is a Dual Core system with 2Gb of Ram.

[Snowman 6008]

Melmo,

According to the help n =  2^m where n is the number of samples, and the data type in is a double, so it can take up to 2²³ (for power 2 quantities) worth of doubles. I've never tried using this function for any samples greater than than 2²³, but you're saying that this works on your system with  2²⁵samples and you get no errors? Based on this segment from the help:

for m = 1, 2,..., 23, the PowerSpectrum method computes the fast Fourier transform of a real-valued sequence using the split-radix algorithm and efficiently scales the magnitude square. The largest power spectrum that the method can compute using the FFT is 

The algorithm MAY only take the first 23 powers, is there any way you can validate that the powersepctrum is correctly processing all of your values when you are using 25 powers?