h a l f b a k e r y
Tempus fudge-it.

meta:

account: browse anonymously, or get an account and write.

 user: pass:
register,

# Venturi-Effect Random Seed Generator

A mechanism for generating natural random numbers
 (+2) [vote for, against]

If you have a fishtank, you see it all the time. You either have a pump or a powerhead that sends bubbles through the water. A "powerhead" is basically a bubble jet, like what you might find in a hot tub. It's about the size of a baseball, and hangs over the edge of a fishtank.

The Venturi Effect is what allows a powerhead to blow bubbles underwater. Basically, a little airtube dangles in the way of the outgoing water jet, and the force of the water leaving the pump sort of "pulls in" the air from inside the tube, causing a pretty big jet of bubbles to shoot out. The rate/amount of bubbles is controlled by a little plastic cap that regulates how much air is allowed to enter the tube on the other end

Long story short, hooking up a high-resolution webcamera so that it looks head-on into the bubble yet creates an absolutely insane amount of natural entropy. I've heard of people using lava lights as a natural entopy source for random numbers (the flow of the wax in the bulb has entropy) but a bubble jet webcam strikes me as as being way, way better for this sort of task. It not only generates heavier entropy, but it makes S-prediction essentially impossible. It's plausible to assume that in the future, things like the fluid dynamics of a lava lamp could be successfully modeled in realtime. However, with this method, you have literally tens of thousands of bubbles of indeterminant size and volume flying toward the camera per second, and all of them are travelling at unusual and unpredictable trajectories -- The ability to accurately model a system like would be next to impossible, since we lack the on-the-spot ability to capture and suspend each individual bubble in the water and measure it's volume, position, and trajectory....and even if we managed to do this, we'd additionally have to have the ability to do this for tens of thousands of bubbles per second. And even if we managed to do all of this, some bubbles are directly absorbed in the water, making measurement and tracking impossible.

 — Bowie23, Mar 04 2004

Fourmilab Hotbits http://www.fourmilab.ch/hotbits/how.html
Describes the process of using radioactive decay for random number generation. Interesting because of the methods it uses to try and reduce bias in the output. [kropotkin, Oct 04 2004]

 Snowflakes can be back-engineered. There's only so many molecules worth of H2O (and various atmospheric impurities) that can be present in a snowflake (i.e. you cant have an infinitely large snowflake, or an infinitely large clump of atmospheric impurity for it to grow from)..coupled with the fact that all snowflakes have 6 sides, which limits the total number of possible permutations of snowflake designs. :( Sure, there are practically trillions of different possible configurations, but the number nonetheless finite.

 White noise isn't as good as it looks, as far as entropy goes. There are non-natural signals present in most white noise, and you can't tell which ones do (and which ones don't) in advance. :(

Finding non-natural signals in a venturi-effect random seed generator would be akin to seeing proof of God. :)
 — Bowie23, Mar 04 2004

I've always thought that the best "easy" random generator might be reading a strip of bits of a random router on the web.
 — theircompetitor, Mar 04 2004

Even if we exclude parity data, the stuff that travels through a router is usually patterned in some form. :(
 — Bowie23, Mar 04 2004

Hmmm -- in the end a seed is a string of bits, right? A string of bits going through a router, this post or the porn image I'm downloading in a different window, how is that "patterned"?
 — theircompetitor, Mar 04 2004

I don't think that we're that good at asking random questions. But I do think sampling from traffic flow would work
 — theircompetitor, Mar 05 2004

As humanbean points out, sampling thermal noise in semiconductors is the most common source of random numbers in hardware generators. However, radiation decay can also be used; I've attached a link to Fourmilab's description of how they use radioactive decay because of the normalisation procedure they outline.
 — kropotkin, Mar 05 2004

 [annotate]

back: main index