Re: 18340: Project 4 Check Node Algorithm

[Shawn Blanton]

Lance

Email the entire staff with all of your questions.
With that said, someone will get back to you shortly.

On 4/24/2013 2:25 PM, Lance Christian wrote:
> Hello,
>
> I have some confusion related to the check node algorithm. I
> originally read the FAQ clarification that we should only compute one
> min value and assign that as the magnitude for every output and the
> sign would be the XOR of all the sign bits which caused every output
> to be identical. After talking with another student about the
> algorithm, I believe I was supposed to have the same magnitude but
> each output should have a sign that is dependent on the XOR of the
> other nodes' sign bits. This could result in some of the outputs
> being the negative of the min magnitude and some being the positive of
> the min magnitudes. Is this correct? Thanks!
>
> -Lance Christian

[Matthew Beckler]

Hi Lance, thanks for the email.

You are correct in that each output of the check node will have the same
minimum magnitude. However, you are not quite correct in saying "the
sign would be the XOR of all the sign bits".

Remember that the overall check node function is that the XOR of all
sign bits is equal to zero. If we write this as an equation, we would have:

X1 ^ X2 ^ X3 ... ^ XN = 0

We can rearrange this equation as such:

X2 ^ X3 ... ^ XN = X1

This shows us how we can compute the correct output sign for each value
(for this example, for X1). So yes, your friend is correct, each output

should have a sign that is dependent on the XOR of the other nodes' sign
bits.

Good luck!

-Matthew