TMS320C5402 : a simple question about delay

Hi everybody,

Here is another simple question for you : I'd like to use a delay
function which would allow me to make delays of x ns. First, is it
feasible ? What's the minimum delay i can do with the TMS320C5402
working with a 100 MHz CPU clock?
I'm wondering : if I write this instruction " for(i=1;i>0;i--); " how
much time does it take?

Once more time, thanks everybody!

Jerome

jerome_lapeyre_mirande@hotmail.com (french_student) writes:

> Hi everybody,
> 
> Here is another simple question for you : I'd like to use a delay
> function which would allow me to make delays of x ns. First, is it
> feasible ? What's the minimum delay i can do with the TMS320C5402
> working with a 100 MHz CPU clock?
> I'm wondering : if I write this instruction " for(i=1;i>0;i--); " how
> much time does it take?
> 
> Once more time, thanks everybody!
> 
> Jerome

Jerome,

If your delay time is a multiple of the sample period, then you can
simply buffer the samples. If it is not, then you can use a fractional
delay filter, which is basically an all-pass filter with a group delay
equal to the desired time delay.
-- 
Randy Yates
Sony Ericsson Mobile Communications
Research Triangle Park, NC, USA
randy.yates@sonyericsson.com, 919-472-1124

french_student wrote:

> Hi everybody,
> 
> Here is another simple question for you : I'd like to use a delay
> function which would allow me to make delays of x ns. First, is it
> feasible ? What's the minimum delay i can do with the TMS320C5402
> working with a 100 MHz CPU clock?
> I'm wondering : if I write this instruction " for(i=1;i>0;i--); " how
> much time does it take?
> 
> Once more time, thanks everybody!
> 
> Jerome

How much memory can you use for a buffer? Store an incoming sample in
the (circular) buffer, and read it out N cycles later. At 100 MHz, the
delay is 10(N+C) ns, where C is the constant number of overhead cycles.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

On Wed, 26 May 2004 11:14:54 -0400, Jerry Avins <jya@ieee.org> wrote:

>french_student wrote:
>
>> Hi everybody,
>> 
>> Here is another simple question for you : I'd like to use a delay
>> function which would allow me to make delays of x ns. First, is it
>> feasible ? What's the minimum delay i can do with the TMS320C5402
>> working with a 100 MHz CPU clock?
>> I'm wondering : if I write this instruction " for(i=1;i>0;i--); " how
>> much time does it take?
>> 
>> Once more time, thanks everybody!
>> 
>> Jerome
>
>How much memory can you use for a buffer? Store an incoming sample in
>the (circular) buffer, and read it out N cycles later. At 100 MHz, the
>delay is 10(N+C) ns, where C is the constant number of overhead cycles.

   From reading his code snippet, he only wants to waste a small,
precise amount of time by using a do-nothing loop, rather than delay
an actual signal
   FWIW, I've done some stuff with the 5402, and a SWAG at the time of
that countdown-from-1-to-0 loop, running at 100MHz, is about 10 or
20ns. If you really need delays like that, there are better ways to do
it. For short delays, a few nop's or a short loop like you have would
be okay (I'd rather use assembly-code nop's or an assembly-code loop,
as its timing won't change with different compile options, whereas it
might with C code). For much longer delays, it would be better to set
up a timer interrupt and let the processor go off and do something
useful.

>Jerry

-----
http://mindspring.com/~benbradley

Ben Bradley wrote:
>    From reading his code snippet, he only wants to waste a small,
> precise amount of time by using a do-nothing loop, rather than delay
> an actual signal
>    FWIW, I've done some stuff with the 5402, and a SWAG at the time of
> that countdown-from-1-to-0 loop, running at 100MHz, is about 10 or
> 20ns. If you really need delays like that, there are better ways to do
> it. For short delays, a few nop's or a short loop like you have would
> be okay (I'd rather use assembly-code nop's or an assembly-code loop,
> as its timing won't change with different compile options, whereas it
> might with C code). For much longer delays, it would be better to set
> up a timer interrupt and let the processor go off and do something
> useful.

A general bit of unsolicited advice: software delays are almost never 
just waiting for time to pass - they are more often allowing time for an 
event to occur.  It's better to monitor the event and proceed after it 
has occurred than it is to wait until the event has had time to happen.

These days the only reason I ever use software delays is when generating 
reset pulses or when forced to write "bit banging" code (like generating 
I2C pulses using digital I/O lines).  But when ever I DO use a delay, I 
make sure there's not a better way to do it.

-- 
Jim Thomas            Principal Applications Engineer  Bittware, Inc
jthomas@bittware.com  http://www.bittware.com          (703) 779-7770
To understand recursion, one must first understand recursion.

"Randy Yates" <randy.yates@sonyericsson.com> wrote in message
news:xxp1xl7b78b.fsf@usrts005.corpusers.net...
> jerome_lapeyre_mirande@hotmail.com (french_student) writes:
>
> > Hi everybody,
> >
> > Here is another simple question for you : I'd like to use a delay
> > function which would allow me to make delays of x ns. First, is it
> > feasible ? What's the minimum delay i can do with the TMS320C5402
> > working with a 100 MHz CPU clock?
> > I'm wondering : if I write this instruction " for(i=1;i>0;i--); " how
> > much time does it take?
> >
> > Once more time, thanks everybody!
> >
> > Jerome
>
> Jerome,
>
> If your delay time is a multiple of the sample period, then you can
> simply buffer the samples. If it is not, then you can use a fractional
> delay filter, which is basically an all-pass filter with a group delay
> equal to the desired time delay.

Randy
I strongly suspect, the OP wasn't trying to implement a delay line for
signal operations but more on the lines of a x ns delay in software before
performing other actions.
BTW, Jim's recommendations should be read before reading my comments because
they address the broader topic.

To answer the original questions..
Yes, it's possible to make delays of x ns. The minimum delay with a 100 MHz
CPU clock is 10 ns - I don't believe this is DSP dependent. If you are
trying to get delays at the resolution of the processor clock in software
and you care about accuracy - you should re-think your approach.

> > I'm wondering : if I write this instruction " for(i=1;i>0;i--); " how
> > much time does it take?
This depends on your compiler. My recommendation is to benchmark this on
your emulator. If you do want to implement delays using such empty loops, I
recommend you use Macros and inline assembly instructions to get better
accuracy. The best option is to use a timer that I'm sure your DSP contains.

Cheers
Bhaskar

> --
> Randy Yates
> Sony Ericsson Mobile Communications
> Research Triangle Park, NC, USA
> randy.yates@sonyericsson.com, 919-472-1124

"Bhaskar Thiagarajan" <bhaskart@deja.com> writes:

> "Randy Yates" <randy.yates@sonyericsson.com> wrote in message
> news:xxp1xl7b78b.fsf@usrts005.corpusers.net...
>> jerome_lapeyre_mirande@hotmail.com (french_student) writes:
>>
>> > Hi everybody,
>> >
>> > Here is another simple question for you : I'd like to use a delay
>> > function which would allow me to make delays of x ns. First, is it
>> > feasible ? What's the minimum delay i can do with the TMS320C5402
>> > working with a 100 MHz CPU clock?
>> > I'm wondering : if I write this instruction " for(i=1;i>0;i--); " how
>> > much time does it take?
>> >
>> > Once more time, thanks everybody!
>> >
>> > Jerome
>>
>> Jerome,
>>
>> If your delay time is a multiple of the sample period, then you can
>> simply buffer the samples. If it is not, then you can use a fractional
>> delay filter, which is basically an all-pass filter with a group delay
>> equal to the desired time delay.
>
> Randy
> I strongly suspect, the OP wasn't trying to implement a delay line for
> signal operations but more on the lines of a x ns delay in software before
> performing other actions.

Ayup - I missed that. Sorry! (I gotta start READING these posts
before I respond to them...)
-- 
%  Randy Yates                  % "So now it's getting late,
%% Fuquay-Varina, NC            %    and those who hesitate
%%% 919-577-9882                %    got no one..."
%%%% <yates@ieee.org>           % 'Waterfall', *Face The Music*, ELO
http://home.earthlink.net/~yatescr

Hi everybody,

Thanks for your answers.

Apparently i wasn't clear enough. I don't want to add a delay to an
incoming signal, i just want to do a short delay before keeping on the
program.
As my prog is in C, and as i'd like it to remain if possible for the
moment,
i 'd rather use a C instruction. So you say a classic loop isn't
accurate and you advice to use the timer. But is it hard to configure?
What would be the instructions? Do i really need to use the timer?

You also say the duration of the classic loop "for(...)" depends on
the compiler. Could you explain why? I thought it only depended on the
DSP's CPU clock.

I think that's all for this subject. But don't worry i have other
questions!! ;:)

Thanks everyone.

Jerome

french_student wrote:

> Hi everybody,
> 
> Thanks for your answers.
> 
> Apparently i wasn't clear enough. I don't want to add a delay to an
> incoming signal, i just want to do a short delay before keeping on the
> program.
> As my prog is in C, and as i'd like it to remain if possible for the
> moment,
> i'd rather use a C instruction. So you say a classic loop isn't
> accurate and you advice to use the timer. But is it hard to configure?
> What would be the instructions?

You need to read the manual to get an understanding of the processor
architecture. With that, the rest is easy.

> Do i really need to use the timer?

No, but you need to read the assembler output each time to see what it
does.

> You also say the duration of the classic loop "for(...)" depends on
> the compiler. Could you explain why? I thought it only depended on the
> DSP's CPU clock.

Different compilers and different optimization levels with the same
compiler can translate the same C expression to different instruction
sequences. A good optimizing compiler might simply eliminate an empty
loop altogether.

> I think that's all for this subject. But don't worry i have other
> questions!! ;:)

Don't worry; we have other answers. Here's one: you can write your delay
loop in assembler, then call that from C. With a little thought, you can
write it to accept a passed delay argument. By determining the actual
delay when the argument is zero, you can compensate for that minimum
time and have the routine delay the actual number of clock cycles you
call for. There will be a minimum, of course, but you will know it.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Hi everybody,

Thanks for all your answers. It's really interesting.

For the moment, i think i don't have to use a delay. But if one day
i'll need it, i'll try the things you advised.

Thanks again.

Jerome


 
> > Hi everybody,
> > 
> > Thanks for your answers.
> > 
> > Apparently i wasn't clear enough. I don't want to add a delay to an
> > incoming signal, i just want to do a short delay before keeping on the
> > program.
> > As my prog is in C, and as i'd like it to remain if possible for the
> > moment,
> > i'd rather use a C instruction. So you say a classic loop isn't
> > accurate and you advice to use the timer. But is it hard to configure?
> > What would be the instructions?
> 
> You need to read the manual to get an understanding of the processor
> architecture. With that, the rest is easy.
> 
> > Do i really need to use the timer?
> 
> No, but you need to read the assembler output each time to see what it
> does.
> 
> > You also say the duration of the classic loop "for(...)" depends on
> > the compiler. Could you explain why? I thought it only depended on the
> > DSP's CPU clock.
> 
> Different compilers and different optimization levels with the same
> compiler can translate the same C expression to different instruction
> sequences. A good optimizing compiler might simply eliminate an empty
> loop altogether.
> 
> > I think that's all for this subject. But don't worry i have other
> > questions!! ;:)
> 
> Don't worry; we have other answers. Here's one: you can write your delay
> loop in assembler, then call that from C. With a little thought, you can
> write it to accept a passed delay argument. By determining the actual
> delay when the argument is zero, you can compensate for that minimum
> time and have the routine delay the actual number of clock cycles you
> call for. There will be a minimum, of course, but you will know it.
> 
> Jerry