relationship between no of bits and roundoff error

As it is clear that if we represent our number by large no of bits,
roundoff error will decrease. Is there any specific relationship between
no of bits and roundoff error?
waiting for your response
Thanks 
---------------------------------------
Posted through http://www.DSPRelated.com

"iarif" <108405@DSPRelated> writes:

> As it is clear that if we represent our number by large no of bits,
> roundoff error will decrease. Is there any specific relationship between
> no of bits and roundoff error?
> waiting for your response

+/- 1/2 LSB, very much like round-off error of a decimal number
at digit n is +/- 10^n / 2 (e.g., for n = 0, +/- 0.5).
-- 
Randy Yates
Digital Signal Labs
http://www.digitalsignallabs.com