back to list

HE Gaussian has a standard deviation of 1%...?

🔗Mike Battaglia <battaglia01@gmail.com>

11/3/2010 9:49:59 PM

Hey all,

A quick question: the canonical HE Gaussian curve has a standard
deviation of "1%". What exactly does that mean? How many cents? I
assume it's not .01 cents.

Sincerely,
Confused

🔗Carl Lumma <carl@lumma.org>

11/3/2010 10:15:28 PM

I think it's 17 cents but you should ask Paul. -C.

At 09:49 PM 11/3/2010, you wrote:
>Hey all,
>
>A quick question: the canonical HE Gaussian curve has a standard
>deviation of "1%". What exactly does that mean? How many cents? I
>assume it's not .01 cents.
>
>Sincerely,
>Confused
>
>

🔗Mike Battaglia <battaglia01@gmail.com>

11/4/2010 4:07:22 AM

On Thu, Nov 4, 2010 at 1:15 AM, Carl Lumma <carl@lumma.org> wrote:
>
> I think it's 17 cents but you should ask Paul. -C.

Found the answer in an old email from Paul. The idea is: s=1% is 17
cents because you treat 1% as the interval 1.01. So
1200*log(1.01)/log(2) = 17.226 cents.

So to convert s to cents, where s is a percentage from 0-100% and
hence from 0-1, the formula is cents = 1200*log(1 + s)/log(2). And to
convert cents to s, the formula is cents*log(2)/1200

s = 2^(cents/1200) - 1

-Mike

🔗Carl Lumma <carl@lumma.org>

11/4/2010 7:05:22 PM

Mike wrote:

>> I think it's 17 cents but you should ask Paul. -C.
>
>Found the answer in an old email from Paul. The idea is: s=1% is 17
>cents because you treat 1% as the interval 1.01. So
>1200*log(1.01)/log(2) = 17.226 cents.

Miraculous. I just pulled 17 cents out of the air. j/k

-Carl