Post by thread:RANDOM NUMBERS....PICing the right number 101

Getting a feel for my tools. ? A.Is it possible to generate RANDOM equally weighted numbers (0-255)on a pic and any simple solutions? ? B.Triangulation....any low resolution(8 or 16 bit answer)sample codes around for study? ? C.Is it unreasonable to crunch numbers and still have time to preform other functions on a pic....trying to feel the weight of my tool....when do I pull in a co-processor or switch micros? Trying to be a better craftsman, STEVE TOMES ____________________________________________________________________ More than just email--Get your FREE Netscape WebMail account today at http://home.netscape.com/netcenter/mail

> Getting a feel for my tools. > ? A.Is it possible to generate RANDOM equally weighted numbers > (0-255)on a pic and any simple solutions? PN generators are easy to implement, it kind of depends on what you mean by random. a 38 bit maximal length PN generator would be pretty hard to predict. The only number you don't get is all 38 bits=0, but if you take a 16 or 32 bit sample, you'll get all possible values in equal probability.

On Fri, 13 Nov 1998, Dave VanHorn wrote: > > Getting a feel for my tools. > > ? A.Is it possible to generate RANDOM equally weighted numbers > > (0-255)on a pic and any simple solutions? > > PN generators are easy to implement, it kind of depends on what you mean > by random. a 38 bit maximal length PN generator would be pretty hard to ANY N bit maximal length generator is EASY to predict because it is guaranteed that each output combination occurs exactly once in a complete cycle thus you know the next one immediately. Of course you also need to tell the interested party that it is 38 bit long and complete. There aren't that many algorythms that an interested party can't try them all out while you sleep on your laurels. > predict. The only number you don't get is all 38 bits=0, but if you take > a 16 or 32 bit sample, you'll get all possible values in equal > probability. Hrrmph. See above. Ok, it takes a determined attacker to do it, but it also depends on what is being protected against whom. Non-maximal and sub-optimal PRNGs are far better at staying unknown. For example two different 31 bit PRNG algorythms operating alternately on the same register and using a certain fraction of the input as output is better imho. It is far more difficult to do probability calculations on this system than on the previous one. I think that thi is a distinct advantage ;) Peter

"Peter L. Peres" wrote: > > ANY N bit maximal length generator is EASY to predict because it is > guaranteed that each output combination occurs exactly once in a complete > cycle thus you know the next one immediately. Of course you also need to > tell the interested party that it is 38 bit long and complete. There > aren't that many algorythms that an interested party can't try them all > out while you sleep on your laurels. But for a sample of bits shorter that the length of the generator, the value will occur more than once, and in ALL equal and longer implementations of the generator. You didn't say what sort of randomity you wanted. For total non-predictability, my favourite is an alpha source and detector. AFAIK, it's purely gaussian, and totally unpredictable. Since it's alphas, all you need for effective shielding is a paper enclosure. > Hrrmph. See above. Ok, it takes a determined attacker to do it, but it > also depends on what is being protected against whom. Who said you have to take your subsample from sequential bits? There's a lot of ways to skin that cat, but first we need to know which cat!