Compression - Re: how to efficiently storing floating point numbers with limited

You are viewing an archived Text-only version of the thread. To view this thread in it's original format and/or if you want to reply to this thread please [click here]

Author

Re: how to efficiently storing floating point numbers with limited

John Reiser

2008-04-01, 6:58 pm

> is there a 16 bit or 24 bit floating point scheme?

Round the IEEE754 representation to (16 - 9) or (24 - 9)
significant [and represented] bits in the fraction.
Convert to full IEEE754 32-bit width for processing.
15 (+1) bits of fraction is enough to represent data from
many physical sensors. 7 (+1) bits probably is poor:
only slightly more than two significant decimal digits.

Scaled fixed point representations such as "16.16" or "14.18"
have been used in some signal-processing environments.

--

Sponsored Links