Prev: Synplify synthesis error
Next: Quartus II under Windows7?
From: onkars on 21 Apr 2010 16:03
I am a student who is trying to model a parallel hardware architecture for
FFT using a C. My aim is to verify the correctness of my architecture and
also estimate the noise introduced when fixed point is used.

Is there any tutorial/book or any help that can guide me in this process of
C modelling --- and especially for fixed point models?



Thank you.

---------------------------------------
Posted through http://www.FPGARelated.com
From: pbljung on 21 Apr 2010 17:17
> I am a student who is trying to model a parallel hardware architecture for
> FFT using a C. My aim is to verify the correctness of my architecture and
> also estimate the noise introduced when fixed point is used.
>
> Is there any tutorial/book or any help that can guide me in this process of
> C modelling --- and especially for fixed point models?

One alternative is to code your FFT in the high-level concurrent
language Mobius which supports parameterized integers, fixed point and
floating point. The Mobius compiler generates synthesizeable Verilog
or VHDL with excellent QoR. On www.codetronix.com there are several
FFT demos including Cooley, combinatorial and r2^2sdf architectures
demonstrating bit-accurate TLM simulations and synthesis. You could
use these as a basis and vary the bitsizes to experimentally observe
quantization noise.

/Per
From: Rob Gaddi on 21 Apr 2010 17:38
On 4/21/2010 1:03 PM, onkars wrote:
> I am a student who is trying to model a parallel hardware architecture for
> FFT using a C. My aim is to verify the correctness of my architecture and
> also estimate the noise introduced when fixed point is used.
>
> Is there any tutorial/book or any help that can guide me in this process of
> C modelling --- and especially for fixed point models?
>
>
>
> Thank you.
>
> ---------------------------------------
> Posted through http://www.FPGARelated.com

You're going to have to write it in VHDL or Verilog eventually anyway.
Might as well do the modeling there too. You could start by writing
purely behavioral code for it, and then have a pretty straightforward
path to making something synthesizable out of it.

--
Rob Gaddi, Highland Technology
Email address is currently out of order
From: Jonathan Bromley on 21 Apr 2010 17:44
On Wed, 21 Apr 2010 15:03:42 -0500, "onkars" wrote:

>I am a student who is trying to model a parallel hardware architecture for
>FFT using a C. My aim is to verify the correctness of my architecture and
>also estimate the noise introduced when fixed point is used.
>
>Is there any tutorial/book or any help that can guide me in this process of
>C modelling --- and especially for fixed point models?

If you download the SystemC-2.0.1 class library from
www.systemc.org you will find a comprehensive
package of template classes for modelling
fixed-point values in C++. You don't have to
do the full SystemC thing to use it, and in
any case the code should give you some
useful ideas.

However, folk who do this kind of stuff all the
time tend to use Matlab, don't they?
--
Jonathan Bromley
From: Rob Gaddi on 21 Apr 2010 17:55
On 4/21/2010 2:38 PM, Rob Gaddi wrote:
> On 4/21/2010 1:03 PM, onkars wrote:
>> I am a student who is trying to model a parallel hardware architecture
>> for
>> FFT using a C. My aim is to verify the correctness of my architecture and
>> also estimate the noise introduced when fixed point is used.
>>
>> Is there any tutorial/book or any help that can guide me in this
>> process of
>> C modelling --- and especially for fixed point models?
>>
>>
>>
>> Thank you.
>>
>> ---------------------------------------
>> Posted through http://www.FPGARelated.com
>
> You're going to have to write it in VHDL or Verilog eventually anyway.
> Might as well do the modeling there too. You could start by writing
> purely behavioral code for it, and then have a pretty straightforward
> path to making something synthesizable out of it.
>

And goddammit, I just got to reading comp.dsp.

http://www.blakjak.demon.co.uk/mul_crss.htm

Go learn that before you use the Internet for anything again. Ever.

--
Rob Gaddi, Highland Technology
Email address is currently out of order
 | 
Pages: 1
Prev: Synplify synthesis error
Next: Quartus II under Windows7?