error in modelsim simulation

V

viswanath

Guest
Hi,
I was trying to implement digital sine modulation in vhdl and for the
same I was multiplying a sine value( which I assumed to be a real
number). For reasons unknown to me though it appears to be an ok
waveform I get error messages stating the following:
Error: XLOCAL <= 0.0 after reduction in SIN(X)
# Time: 479952500 ps Iteration: 0 Instance: /sigtst/s1
At this instant of time the value of sine is of the order 1e -14,
1e-15, 1e-13.
The code at which I think is the problem is :
sin_theta := sin(math_2_pi*real(now/1 ps)/real(period/ 1 ps));
What could that problem be? and how can one fix that?
I need this because for a further multiplication with another real
number I am getting a FATAL error message and the simulation stops. It
says the real number range has exceeded though I am multiplying two
real numbers which are within that range one the above number and
another number which is also a sine value. Though this cannot exceed
the range I get that message.
I would greatly appreciate any help and suggestions.
Thanking You in Anticipation,
Yours Truly,
Viswanath Daita
 
On 21 Apr 2004 21:25:40 -0700, daita@eng.usf.edu (viswanath) wrote:

Hi,
I was trying to implement digital sine modulation in vhdl and for the
same I was multiplying a sine value( which I assumed to be a real
number). For reasons unknown to me though it appears to be an ok
waveform I get error messages stating the following:
Error: XLOCAL <= 0.0 after reduction in SIN(X)
# Time: 479952500 ps Iteration: 0 Instance: /sigtst/s1
At this instant of time the value of sine is of the order 1e -14,
1e-15, 1e-13.
The code at which I think is the problem is :
sin_theta := sin(math_2_pi*real(now/1 ps)/real(period/ 1 ps));
What could that problem be? and how can one fix that?
Click to expand...
I think the problem is internal to the SIN() function. You are
asking it to calculate the sine of a very large angle, far larger
than one rotation. To perform this calculation, the SIN()
function needs to reduce the angle so that it is within some
range (probably 0 to pi/2, though I'm not sure exactly how
the SIN() function works internally). My guess is that
there is some limitation to the way this has been implemented,
and you have discovered a corner-case bug.

If my guess is right, the obvious solution is to do the
reduction yourself:

sin_theta := sin(math_2_pi * real( (now/1 ps) mod (period/1 ps) ));

The IEEE standard specifies that accuracy may be degraded for
very large input angles, but doesn't permit failure like this.
Consequently it would be a good idea to capture the input value
(479952500/period) and report it to support@model.com .
--
Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how
VHDL, Verilog, SystemC, Perl, Tcl/Tk, Verification, Project Services

Doulos Ltd. Church Hatch, 22 Market Place, Ringwood, BH24 1AW, UK
Tel: +44 (0)1425 471223 mail:jonathan.bromley@doulos.com
Fax: +44 (0)1425 471573 Web: http://www.doulos.com

The contents of this message may contain personal views which
are not the views of Doulos Ltd., unless specifically stated.
 
Hi Jonathan,
Thanks for the reply.Though it sounds exciting that I have discovered
a bug in modelsim I need a way to get around this problem.
I didnot explain to you what I did previously. I am trying to model a
continuous sine wave and the expression I have in the previous mail is
what I am using to calculate the modulation.((const)* sin(wt) )
How could one implement the same instead of using the kind of
expression that I used. where the omega frequency is in the order of
Mhz.
Can you please let me know?
Looking forward to your reply.
Thanking You,
Viswanath(aka Vishy)

Jonathan Bromley <jonathan.bromley@doulos.com> wrote in message news:<la0f809ap24rpc3fcu172enek2srgd4jau@4ax.com>...
On 21 Apr 2004 21:25:40 -0700, daita@eng.usf.edu (viswanath) wrote:

Hi,
I was trying to implement digital sine modulation in vhdl and for the
same I was multiplying a sine value( which I assumed to be a real
number). For reasons unknown to me though it appears to be an ok
waveform I get error messages stating the following:
Error: XLOCAL <= 0.0 after reduction in SIN(X)
# Time: 479952500 ps Iteration: 0 Instance: /sigtst/s1
At this instant of time the value of sine is of the order 1e -14,
1e-15, 1e-13.
The code at which I think is the problem is :
sin_theta := sin(math_2_pi*real(now/1 ps)/real(period/ 1 ps));
What could that problem be? and how can one fix that?

I think the problem is internal to the SIN() function. You are
asking it to calculate the sine of a very large angle, far larger
than one rotation. To perform this calculation, the SIN()
function needs to reduce the angle so that it is within some
range (probably 0 to pi/2, though I'm not sure exactly how
the SIN() function works internally). My guess is that
there is some limitation to the way this has been implemented,
and you have discovered a corner-case bug.

If my guess is right, the obvious solution is to do the
reduction yourself:

sin_theta := sin(math_2_pi * real( (now/1 ps) mod (period/1 ps) ));

The IEEE standard specifies that accuracy may be degraded for
very large input angles, but doesn't permit failure like this.
Consequently it would be a good idea to capture the input value
(479952500/period) and report it to support@model.com .
--
Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how
VHDL, Verilog, SystemC, Perl, Tcl/Tk, Verification, Project Services

Doulos Ltd. Church Hatch, 22 Market Place, Ringwood, BH24 1AW, UK
Tel: +44 (0)1425 471223 mail:jonathan.bromley@doulos.com
Fax: +44 (0)1425 471573 Web: http://www.doulos.com

The contents of this message may contain personal views which
are not the views of Doulos Ltd., unless specifically stated.
Click to expand...
 
On 22 Apr 2004 11:24:35 -0700, daita@eng.usf.edu (viswanath) wrote:

Hi Jonathan,
Thanks for the reply.Though it sounds exciting that I have discovered
a bug in modelsim I need a way to get around this problem.
Click to expand...
Read my post again. I gave you a solution.

I didnot explain to you what I did previously. I am trying to model a
continuous sine wave and the expression I have in the previous mail is
what I am using to calculate the modulation.((const)* sin(wt) )
Click to expand...
Yup. Your problem, of course, is that (omega.t) grows without
(theoretical) limit as the simulation runs. I suggested one of
many possible ways to work around that.

How could one implement the same instead of using the kind of
expression that I used.
Click to expand...
Just in case you missed it last time:

sin_theta := sin(math_2_pi * real( (now/1 ps) mod (period/1 ps) ));
--
Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how
VHDL, Verilog, SystemC, Perl, Tcl/Tk, Verification, Project Services

Doulos Ltd. Church Hatch, 22 Market Place, Ringwood, BH24 1AW, UK
Tel: +44 (0)1425 471223 mail:jonathan.bromley@doulos.com
Fax: +44 (0)1425 471573 Web: http://www.doulos.com

The contents of this message may contain personal views which
are not the views of Doulos Ltd., unless specifically stated.
 
Hi Jonathan,
Thanks for your reply. I could get this thing finally working.
Viswanath

Jonathan Bromley <jonathan.bromley@doulos.com> wrote in message news:<f6hh80hhr5rs3fat9po2cifn4orcf2hh2m@4ax.com>...
On 22 Apr 2004 11:24:35 -0700, daita@eng.usf.edu (viswanath) wrote:

Hi Jonathan,
Thanks for the reply.Though it sounds exciting that I have discovered
a bug in modelsim I need a way to get around this problem.

Read my post again. I gave you a solution.

I didnot explain to you what I did previously. I am trying to model a
continuous sine wave and the expression I have in the previous mail is
what I am using to calculate the modulation.((const)* sin(wt) )

Yup. Your problem, of course, is that (omega.t) grows without
(theoretical) limit as the simulation runs. I suggested one of
many possible ways to work around that.

How could one implement the same instead of using the kind of
expression that I used.

Just in case you missed it last time:

sin_theta := sin(math_2_pi * real( (now/1 ps) mod (period/1 ps) ));
--
Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how
VHDL, Verilog, SystemC, Perl, Tcl/Tk, Verification, Project Services

Doulos Ltd. Church Hatch, 22 Market Place, Ringwood, BH24 1AW, UK
Tel: +44 (0)1425 471223 mail:jonathan.bromley@doulos.com
Fax: +44 (0)1425 471573 Web: http://www.doulos.com

The contents of this message may contain personal views which
are not the views of Doulos Ltd., unless specifically stated.
Click to expand...
 
You must log in or register to reply here.

Welcome to EDABoard.com

Sponsor

Back
Top