Greetings all,
I apologize in advance if this topic has been addressed, I couldn't find anything after searching.
I have a fixed point number <+/- 13,13> which goes from -4096 to 4095. I would like to divide that by 2048, which happens to be a very simple bit shift operation. If you keep the word and shift the q-point to cast it as <13,2>, I get the answer I'm looking for.
This works outside of an FPGA, but the type-casting block is not FPGA-compatible.
The 'logical shift' block help documentation says to use the 'scale by power of 2' block with signed data types, however when fed with a <13,13> fxp source it outputs a <13,13> fxp, so I just lose all of the fractional bits.
Is there some way to re-cast (and be compatible with FPGA) the fixed point variable such that the word isn't changed but the virtual q-point gets moved? I can accomplish it with a 'high throughput divide' block but that seems like overkill for a simple shift scaling.
This is a super simple / common operation so I must be missing something. In other languages this would be trivial. Any thoughts?
Thanks in advance,
Chris
EDIT:
About 5 minutes after posting the question, I figured it out. Isn't that the way it always goes?
Anyways, for those that see this when looking in the future, I used a 'convert number to boolean array' connected to a 'convert boolean array to number' blocks, and you can set the output format of the 'convert boolean array to number' block to whatever fixed point format the heart desires.
