PLL on TMS320C6713 to generate 1Hz output

paryanz wrote:
> In that thread we were not considering the implementation using C6713 or
> some kind of DSP processor and the project requirements were not very clear
> to me.
> 
> Now I know for sure what hardware I have and what I need to generate, If
> you have any specific advice with respect to implementation on a processor
> to generate the 1 Hz output with those freqs as a reference input. I would
> really appreciate your help.

I'm not holding back. I know how I would start to do it, but I haven't 
actually done that task so I have no advice worth giving.

My comment about saying what you mean applied to the frequency you 
cited. When you write 10 Hz accurate to GPS standards, a reader can't 
know that you might mean 10.01 Hz.

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

paryanz wrote:

> In that thread we were not considering the implementation using C6713 or
> some kind of DSP processor and the project requirements were not very clear
> to me.
> 
> Now I know for sure what hardware I have and what I need to generate, If
> you have any specific advice with respect to implementation on a processor
> to generate the 1 Hz output with those freqs as a reference input.

Here is the specific advice: hire a consultant. In this newsgroup, at 
least ten persons can fix your problem at no time.

> I would really appreciate your help.

It depends. How much is the real appreciacion?

VLV

Vladimir Vassilevsky schrieb:
> 
> Here is the specific advice: hire a consultant. In this newsgroup, at 
> least ten persons can fix your problem at no time.
> 

VV, greedy again, eh? Hire a consulatant for such a simple problem?



Here's my advice:

Connect your 10Hz signal to a GPIO pin. Install an interrupt handler + 
event for that GPIO pin.

Then inside the interrupt handler: Increment a counter. If you reach 10 
toggle the state of another GPIO pin and reset your counter.

You'll get a pretty good 1Hz reference that way. Your accuracy is 
limited by the interrupt thresold (the guys who take care about the 
compiler settings should know).

Assuming that you have GPIO pins free the entire thing will cost around 
500 bytes of code and will have nearly no influence on the system 
performance, e.g. it's a free meal.

Cheers,
   Nils

Nils wrote:

> Vladimir Vassilevsky schrieb:
> 
>>
>> Here is the specific advice: hire a consultant. In this newsgroup, at 
>> least ten persons can fix your problem at no time.
>>
> 
> VV, greedy again, eh?

The 101 computer science: Lamer must pay.

> Hire a consulatant for such a simple problem?

Don't you see? Explain or not, he will never be able to do that. At 
least not in a year or so. Want a bet?

> Here's my advice:
> 
> Connect your 10Hz signal to a GPIO pin. Install an interrupt handler + 
> event for that GPIO pin.

Next question: what is GPIO and how to install the interrupt handler...

> Then inside the interrupt handler: Increment a counter. If you reach 10 
> toggle the state of another GPIO pin and reset your counter.
> 
> You'll get a pretty good 1Hz reference that way.

You'll get a pretty bad 1Hz reference with the interrupt latency jitter.

Hint: the 6713 timers can be clocked externally.

> Your accuracy is 
> limited by the interrupt thresold (the guys who take care about the 
> compiler settings should know).

:))))))


> Assuming that you have GPIO pins free the entire thing will cost around 
> 500 bytes of code and will have nearly no influence on the system 
> performance, e.g. it's a free meal.

Naive.


VLV

 > The 101 computer science: Lamer must pay.

There are very little real lamers out there but a lot of guys who jus 
don't have the experience..

 > Next question: what is GPIO and how to install the interrupt
 > handler...

Dsp-Bios documentation along with the example codes will tell and teach 
them.

 > You'll get a pretty bad 1Hz reference with the interrupt latency
 > jitter.

That depends...

 >> Assuming that you have GPIO pins free the entire thing will cost
 >> around 500 bytes of code and will have nearly no influence on the
 >> system performance, e.g. it's a free meal.
 >
 > Naive.

Not really.

It depends on what you want to do with the clock, and how many wobble 
and jitter you an tolerate.

Little example from germany:

The 50hz cylce from the power supply has lots of jitter and wobble, but 
over a longer timespan (say a day or two) they are a rock-solid time 
reference.

Same goes to the 18khz pilot-tone for radio FM-modulation.

In germany both frequencies are pll'ed from the DCF77 timebase. That 
means we have a caesium atomic clock as a refernce. Sure - it can be off 
for a second once in a while, but over a week such differences will even 
out.

I'm pretty sure the same is done with such frequencies all over europe 
at last.



Back to the topic:

What would you use a clock at one Hz for that can't live with a jitter 
of 2 or 4 milliseconds?

Nils wrote:


> 
> Back to the topic:
> 
> What would you use a clock at one Hz for that can't live with a jitter 
> of 2 or 4 milliseconds?

There is a whole lot of different applications where you have to 
synchronize multiple channels. Ranging from pro audio to active 
beamforming.

BTW not too long ago I developed the data acquisition board. This board 
can be synchronized to the GPS 1pps reference, and/or used as the clock 
master for the other data ascqusition boards. The short term jitter spec 
was 1e-8; we actually did about 1e-9.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

Thank you Nils for your advice and direction....