浮点转换 [英] Floating point Conversion

问题描述

你好

我有一个关于数字系统的问题，我只是一个正在学习不同信号/图像处理领域的人，这个问题是

（
Hello

I have a question about number systems and i am just a begginer who is learning different field of signal/image processing the wuestion is

(􀀀1)s  (1 + f )  2^(e-eb) :
The format has a word length of q bits, the length of the significand part representing
f is k bits, and the bias is eb (emax above is the maximum integer representable
with the bits allocated to the exponent).1
Let q=12 k=5 and eb=15, and determine the
word length p and the fraction length n of a fixed-point format (binary point scaling)
which can represent all values representable in the specified floating-point
format.?


I am not even sure if this is the right place to ask but anyways i am trying . If some has time to explain it to me or show me how to convert it i mean theoretically than i might code it.

Regards

Thank you so much
解决方案
Do you have a text book or any scripts from your lecture?

Lookup the terms there first, then understand the context of the formula, ask your teacher if you have any unclear items. 



Check the internet for floating point standards, e.g. http://en.wikipedia.org/wiki/IEEE_floating_point[^].

Check what a fixed-point number is, e.g. http://en.wikipedia.org/wiki/Fixed-point_arithmetic[^]. See also floating point: http://en.wikipedia.org/wiki/Floating_point[^].

Cheers

Andi



PS: As a hint: the fixed-point number must be capable to store all values that the floating-point number can store. Find out first, what the maximum floating-point number is. Then the minimum (most negative value if it is about positive *and* negative numbers). Then find the number closest to zero (the smallest, the one with the highest negative exponent and taking in consideration the k bits that are even below that number).



I get roughly calculating something like 70 bits and the fixed point somewhere shy above 50.


                        
这篇关于浮点转换的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！