Typically what is suggested is 10e18 spins for a quality signal to noise ratio.  Therefore, a mole of sample is ample for protons.  Depending on the density of the material and number of protons, it can be anywhere from tens of micrograms to tens of milligrams.  A good way to know how much you need is to back calculate from the volume of the rotor (e.g. a 4 mm rotor has a volume of ~80 uL.  Therefore you can use the density of your sample to figure out how much sample you need).  

As for the 13C, if you pack the rotor fully and run a CP-MAS experiment, you should be able to see the 13C signal easily.  It will probably take some signal averaging but it is a rather routine experiment.    

 For 13Carbon around 25-30 mg and solid state even more. not more than 100 mg.....

Actually, there is no quick answer for your question without detailed descriptions. For example, the proton detection for solid-state NMR is only possible at fast MAS condition (at least >40 kHz) to achieve narrow enough 1H linewidth. But the rotors which can spin higher than 40 kHz are small, which means you are not able to pack a lot of samples in it. So if you plan to run ssNMR on small organic compounds, 1H spectra cannot provide much useful information as solution NMR. For 13C detection, you can use either 4mm (80uL volume) or 3.2mm rotors (30uL volume) which usually can take ~50mg or ~25mg samples respectively. With that amount of sample, you should be able to acquire a 1D spectrum within in a few minutes or a few hours for 13C labeled samples or natural abundant samples respectively.

Thanks All

Very interesting topic. I've never conducted NMR but I was wondering: is it physically possible to get a NMR spectrum from a microarea, as is being done in a SEM microscope or electron microprobe (a "NMR microprobe")?