Fixed-Point binary unpacking in python

Question

I am struggling with a number-format problem in Python 3.6. My goal is to convert binary data from a file into printable decimal numbers. As an example, I need to convert two little-endian bytes in the form of a byte string...

b'\x12\00'

into its big-endian binary form...

0000000000010010

and finally to its 16-bit fixed-point Q15 decimal number form...

(1 / 4096) + (1 / 16384) = 0.00030517578 (basically, we've made the 2 bytes above human-readable)

In my failed attempts, the struct.unpack function seemed promising, but my low-level / number representation experience just isn't very mature at the moment.

Failed Attempt:

struct.unpack('<h', b'\x12\x00') # Yields (18,)

The above code gets me "18", which would be fine if the bytes represented an integer, but they do not.

Any help / advice would be appreciated. Thank you!

The bytes do represent an integer - which has been shifted by 15 bits. Divide by 32768 (2**15) to get the actual Q15 value. (This doesn't match the value you calculated, but that's because you did the math wrong - the two set bits actually have place values of 1/2048 and 1/16384.) — jasonharper
@jasonharper You're right-- nice catch! Your solution works, and I believe I understand your approach. Thank you! — Austin Yarger

Austin Yarger Austin Yarger · Accepted Answer · 2017-06-06T20:13:18

Answered by @jasonharper in the question comments--

The bytes do represent an integer - which has been shifted by 15 bits. Divide by 32768 (2**15) to get the actual Q15 value. (This doesn't match the value you calculated, but that's because you did the math wrong - the two set bits actually have place values of 1/2048 and 1/16384.)

I achieved the proper value via the following code--

struct.unpack('<h', b'\x12\x00')[0] / (2**15)

Fixed-Point binary unpacking in python

1 Answers