-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On Wed, 20 Oct 1999, Clifford Heath wrote:
> It doesn't matter in the slightest how much signal there is in the noise.
Nit - it does if clips out the input.
> What matters is how much entropy there is. Our experiments showed a
> variance (cheap 16bit sound card, microphone input with nothing plugged
> in, full record volume) of around 8-11, or over three bits per sample.
> We're basically looking for quantum shot noise in the input transistors,
> it doesn't make an appreciable difference if you short-circuit the input.
> Differencing the two channels would only reduce the available entropy, which
> holds true until the variance falls below 1 bit per sample - I don't know
> any 16 bit cards with this property though.
> We just use the bottom bit, but could use the bottom three (on this card),
> and could for example have used the parity over the entire 16 bit sample.
> Again, having signal in your noise doesn't hurt any as long as it doesn't
> swamp the noise.
Agreed (I use a hash of all 16 bits in my software), but removing as much
signal as possible allows one to make better estimates of the available
entropy.
Regards,
Damien Miller
- --
| "Bombay is 250ms from New York in the new world order" - Alan Cox
| Damien Miller - http://www.mindrot.org/
| Email: [EMAIL PROTECTED] (home) -or- [EMAIL PROTECTED] (work)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.0 (GNU/Linux)
Comment: For info see http://www.gnupg.org
iD4DBQE4DWJBormJ9RG1dI8RAm7zAJ0Z+UlOWcPNWBvEk/TTTwXmBeWRfgCYj+6E
hKK5NMkj0HfZjqPiweZH2A==
=09kn
-----END PGP SIGNATURE-----
______________________________________________________________________
OpenSSL Project http://www.openssl.org
Development Mailing List [EMAIL PROTECTED]
Automated List Manager [EMAIL PROTECTED]
