Message

Reply | Reply all

Date Prev | Date Next | Thread Prev | Thread Next Date Index | Thread Index

From: Víctor López<victor.lopez@o...>
Date: Wed Sep 19 05:42:31 CEST 2007
Subject: [oc] Who is going to work on cores?
Hello,

You said:
> I have a problem in deciding the format of the input data stream. How
> to set the format while i input the data through Test bench. I am
> using twos complement format to input negative numbers.
> For example: My data word is 16 bits. So how are the bits assigned?
> What I understand is 1 bit is sign bit, one bit for intefer and rest
> 14 bits for fractional value. Is this format correct or is there a way
> to define the input sequence.

So my answer is still the same: Two's compliment does not have anything
to do with fractions, it is just about integers. The proper name for the
format you are talking about is fixed-point numbers. Anyways two's
compliment doesn't really just use one sign bit, the whole final stored
value is affected by the sign, read the link I put on my earlier response.

What are the input formats allowed by the FFT core? Are we talking about
the xilinx coregen one whose datasheet can be found here
http://www.xilinx.com/ipcenter/catalog/logicore/docs/xfft.pdf ?
Then from what I read in the middle of page 3:
"As with unscaled arithmetic, for scaled and block floating point
arithmetic, the core does not have a specific location for the binary
point. The location of the binary point in the output data is inherited
from the input data and then shifted by the scaling applied."

My remembrance of the internals of FFT is somewhat distorted from the
time I learned it, but to me the text reads as: the place where you set
the fixed-point is unimportant, as long as you respect it in all the
inputs, so at the output you can expect it in the same place (unless you
set the scaling parameter of the core to whatever).

When in doubt, the best thing to do is the "try it yourself", get some
decimal values, FFT them in MatLab or some other software to get the
good outputs and then simulate the same inputs in the FFT core to see if
the output matches. I also wish the 46 pages of the datasheet where
clearer on these things.

Regards,

Víctor López

Dhawal Kadao escribió:
> Hi Victor,
>
> For the input to FFT the numbers are less than or equal to 1. for eg
> if the number is 1.5 the binary representatoin is 11 the first 1 being
> the integer and second being the fractional part.
> For the input the decimal point is implicit. I am looking for the
> default location where the core puts the point.
>
> Regards,
> Dhawal.
>
>
> On 9/18/07, *Víctor López* <victor.lopez@o...
> <mailto:victor.lopez@o...>> wrote:
>
> Hi Dhawal,
>
> I really do think you should study more carefully the different
> signed integer notations. For example, just googling for "two's
> compliment" I found this fairly well explained link:
> http://en.wikipedia.org/wiki/Two's_complement
> <http://en.wikipedia.org/wiki/Two%27s_complement>
>
> Take a good look at it and it'll answer your question. By the way,
> integers are integer numbers.... that is: they don't ever have a
> fractional part.
>
> Regards,
>
> Víctor López
>
> Dhawal Kadao escribió:
> >
> > Thanks Balaji,
> >
> > I have a problem in deciding the format of the input data
> stream. How
> > to set the format while i input the data through Test bench. I am
> > using twos complement format to input negative numbers.
> > For example: My data word is 16 bits. So how are the bits assigned?
> > What I understand is 1 bit is sign bit, one bit for intefer and
> rest
> > 14 bits for fractional value. Is this format correct or is there
> a way
> > to define the input sequence.
> >
> > Regards,
> > Dhawal.
> >
> > On 9/15/07, * rbala4ums@h...
> <mailto:rbala4ums@h...> <mailto:rbala4ums@h...
> <mailto:rbala4ums@h...>>*
> > <rbala4ums@h... <mailto:rbala4ums@h...> <mailto:
> rbala4ums@h... <mailto:rbala4ums@h...>>> wrote:
> >
> > Hi Kadao,
> >
> > Xilinx's IP cores are made so easy that it is very much
> explanatory on
> > its own. You can get reference materials on IP cores by doing a > > google > > search like "Xilinx IP Coregen Memory cores" or blah blah > blah.... > > > > if you need info on any particular or specific core, you could > > actually open a Xilinx project in ISE, add the core as a > component, > > and while generating the core using Xilinx's Coregen Wizard, > you will > > actually see a tab or link stating "More info on the core" > if you > > click that, the wizard directly opens the pdf doc about the > specific > > coregen component on a new browser. That is the best way to > get info > > on any particular core. If you need help on instantiating > any core, > > you will have to do a google search on some examples. > > > > There are lots and lots of tutorials written by xilinx, and by > > students of many many research groups from various > universities that > > are openly available on the internet, it just matters on > what kind of > > key words that you use in google. Better you use, more > results you > > get. > > > > As i said FFT's and Memory cores are the most popular cores > that are > > being used by so many people, so it should be easier to > find, also you > > might want to do a search in "Google's Code Search" also, > you will > > find so many examples(though most results are from the > projects of > > open cores) but you will get some good results on that too. > > > > Thanks. > > > > Balaji R > > > > > > ----- Original Message ----- > > From: Dhawal Kadao< dhawal.kadao@g...> > > To: > > Date: Fri Sep 14 16:34:56 CEST 2007 > > Subject: [oc] Who is going to work on cores? > > > > > Asim, > > > > > > I am working on Xilinx ip cores. I have to use fftcore and > memory > > > cores in > > > my project. I need some reference material and help woth > my core. > > > Thanks, > > > Dhawal. > > > On 9/13/07, asim shehzad <attachment-0001.html > > > > > > > > _______________________________________________ > > http://www.opencores.org/mailman/listinfo/cores > > > > > > > ------------------------------------------------------------------------ > > > > _______________________________________________ > > http://www.opencores.org/mailman/listinfo/cores > > _______________________________________________ > http://www.opencores.org/mailman/listinfo/cores > > > ------------------------------------------------------------------------ > > _______________________________________________ > http://www.opencores.org/mailman/listinfo/cores