HELP! VHDL real TYPE produces infinite decimal. How to app

Hello

I am fairly new to VHDL and have hit a wall in my code.

Basically, I am wanting to convert a 50% duty cycle frequency signal into a period with microsecond units. In VHDL, I perform a 1/frequency operation in order to get the period and then convert it to seconds. I declared the period as a real type. The problem I have is when I end up with a frequency that creates an infinite decimal for the period calculation (Example: 1/3 = 0.333333..... )My simulator cannot handle the infinite decimal.

Is there a way to limit the precision of a REAL type number in VHDL? Should I use a floating point? I appreciate the help!

 

[Alan Fitch]

On 04/06/13 18:04, adam.kimura@googlemail.com wrote:

Hello

I am fairly new to VHDL and have hit a wall in my code.

Basically, I am wanting to convert a 50% duty cycle frequency signal into a period with microsecond units. In VHDL, I perform a 1/frequency operation in order to get the period and then convert it to seconds. I declared the period as a real type. The problem I have is when I end up with a frequency that creates an infinite decimal for the period calculation (Example: 1/3 = 0.333333..... )My simulator cannot handle the infinite decimal.

Is there a way to limit the precision of a REAL type number in VHDL? Should I use a floating point? I appreciate the help!

That sounds strange.

Can you post the code and the error message?

To answer you last question, REAL *is* floating point.

regards
Alan

--
Alan Fitch

 

[rickman]

Yes, the REAL data type in VHDL has limited precision just like most
data types in programming languages. I'm not sure what error you might
have seen. I can't think of anything in VHDL that would have a problem
with your use of a REAL data type unless you were trying to synthesize
it. To turn 1/3 into a real number you must first convert the 1 and the
3 into reals and then divide them as real. No, it won't give you
exactly 1/3, but it will not give you an error.

I think the error must have been related to how you were trying to
convert them.

Rick



On 6/4/2013 7:59 PM, adam.kimura@googlemail.com wrote:

Hello Alan,

I think I got it working. I honestly don't know what happened. I was trying to re-create the error that I was getting, but it seemed to work fine now. Is there a limit on the number of decimal places a REAL type can have? From what I understood from the error message, the decimal places were infinite; which caused the simulation error. I understood this to mean that VHDL cannot represent such a number beyond a certain number of decimal places. In languages like C++, a number like 1/3 cannot be accurately represented since it is 0.3333333.... (re-occurring 3.) I have the situation where I am dividing 1 by decimal numbers... hence the long decimals.

Thanks again for your help.

ak


On Tuesday, June 4, 2013 3:18:30 PM UTC-4, Alan Fitch wrote:
On 04/06/13 18:04, adam.kimura@googlemail.com wrote:

Hello



I am fairly new to VHDL and have hit a wall in my code.



Basically, I am wanting to convert a 50% duty cycle frequency signal into a period with microsecond units. In VHDL, I perform a 1/frequency operation in order to get the period and then convert it to seconds. I declared the period as a real type. The problem I have is when I end up with a frequency that creates an infinite decimal for the period calculation (Example: 1/3 = 0.333333..... )My simulator cannot handle the infinite decimal.



Is there a way to limit the precision of a REAL type number in VHDL? Should I use a floating point? I appreciate the help!





That sounds strange.



Can you post the code and the error message?



To answer you last question, REAL *is* floating point.



regards

Alan



--

Alan Fitch

--

Rick

 

Hello Alan,

I think I got it working. I honestly don't know what happened. I was trying to re-create the error that I was getting, but it seemed to work fine now. Is there a limit on the number of decimal places a REAL type can have? From what I understood from the error message, the decimal places were infinite; which caused the simulation error. I understood this to mean that VHDL cannot represent such a number beyond a certain number of decimal places. In languages like C++, a number like 1/3 cannot be accurately represented since it is 0.3333333.... (re-occurring 3.) I have the situation where I am dividing 1 by decimal numbers... hence the long decimals.

Thanks again for your help.

ak


On Tuesday, June 4, 2013 3:18:30 PM UTC-4, Alan Fitch wrote:

On 04/06/13 18:04, adam.kimura@googlemail.com wrote:

Hello



I am fairly new to VHDL and have hit a wall in my code.



Basically, I am wanting to convert a 50% duty cycle frequency signal into a period with microsecond units. In VHDL, I perform a 1/frequency operation in order to get the period and then convert it to seconds. I declared the period as a real type. The problem I have is when I end up with a frequency that creates an infinite decimal for the period calculation (Example: 1/3 = 0.333333..... )My simulator cannot handle the infinite decimal.



Is there a way to limit the precision of a REAL type number in VHDL? Should I use a floating point? I appreciate the help!





That sounds strange.



Can you post the code and the error message?



To answer you last question, REAL *is* floating point.



regards

Alan



--

Alan Fitch

 

Thank you Rick... very helpful information. Thanks for taking the time to respond. I sincerely appreciate it!

Best,

Adam.


On Tuesday, June 4, 2013 9:03:30 PM UTC-4, rickman wrote:

Yes, the REAL data type in VHDL has limited precision just like most

data types in programming languages. I'm not sure what error you might

have seen. I can't think of anything in VHDL that would have a problem

with your use of a REAL data type unless you were trying to synthesize

it. To turn 1/3 into a real number you must first convert the 1 and the

3 into reals and then divide them as real. No, it won't give you

exactly 1/3, but it will not give you an error.



I think the error must have been related to how you were trying to

convert them.



Rick







On 6/4/2013 7:59 PM, adam.kimura@googlemail.com wrote:

Hello Alan,



I think I got it working. I honestly don't know what happened. I was trying to re-create the error that I was getting, but it seemed to work fine now. Is there a limit on the number of decimal places a REAL type can have? From what I understood from the error message, the decimal places were infinite; which caused the simulation error. I understood this to mean that VHDL cannot represent such a number beyond a certain number of decimal places. In languages like C++, a number like 1/3 cannot be accurately represented since it is 0.3333333.... (re-occurring 3.) I have the situation where I am dividing 1 by decimal numbers... hence the long decimals.



Thanks again for your help.



ak





On Tuesday, June 4, 2013 3:18:30 PM UTC-4, Alan Fitch wrote:

On 04/06/13 18:04, adam.kimura@googlemail.com wrote:



Hello







I am fairly new to VHDL and have hit a wall in my code.







Basically, I am wanting to convert a 50% duty cycle frequency signal into a period with microsecond units. In VHDL, I perform a 1/frequency operation in order to get the period and then convert it to seconds. I declared the period as a real type. The problem I have is when I end up with a frequency that creates an infinite decimal for the period calculation (Example: 1/3 = 0.333333..... )My simulator cannot handle the infinite decimal.







Is there a way to limit the precision of a REAL type number in VHDL? Should I use a floating point? I appreciate the help!











That sounds strange.







Can you post the code and the error message?







To answer you last question, REAL *is* floating point.







regards



Alan







--



Alan Fitch







--



Rick

 

[Alan Fitch]

On 05/06/13 00:59, adam.kimura@googlemail.com wrote:

Hello Alan,

I think I got it working. I honestly don't know what happened. I was trying to re-create the error that I was getting, but it seemed to work fine now. Is there a limit on the number of decimal places a REAL type can have? From what I understood from the error message, the decimal places were infinite; which caused the simulation error. I understood this to mean that VHDL cannot represent such a number beyond a certain number of decimal places. In languages like C++, a number like 1/3 cannot be accurately represented since it is 0.3333333.... (re-occurring 3.) I have the situation where I am dividing 1 by decimal numbers... hence the long decimals.

Thanks again for your help.

When I'm teaching VHDL courses, I sometimes joke that we need CCTV on
every keyboard so when people say "oh, it works now!", I could just play
back the tape to find out exactly what they typed

From VHDL 2002 onwards (I think - or was it VHDL 93?) REAL was defined
to be IEEE P754 - in earlier versions of VHDL it was just guaranteed to
have at least a certain range and precision (which I can't recall)

regards
Alan

ak


On Tuesday, June 4, 2013 3:18:30 PM UTC-4, Alan Fitch wrote:
On 04/06/13 18:04, adam.kimura@googlemail.com wrote:

Hello



I am fairly new to VHDL and have hit a wall in my code.



Basically, I am wanting to convert a 50% duty cycle frequency signal into a period with microsecond units. In VHDL, I perform a 1/frequency operation in order to get the period and then convert it to seconds. I declared the period as a real type. The problem I have is when I end up with a frequency that creates an infinite decimal for the period calculation (Example: 1/3 = 0.333333..... )My simulator cannot handle the infinite decimal.



Is there a way to limit the precision of a REAL type number in VHDL? Should I use a floating point? I appreciate the help!





That sounds strange.



Can you post the code and the error message?



To answer you last question, REAL *is* floating point.



regards

Alan



--

Alan Fitch

--
Alan Fitch