>
> Byron A Jeff wrote:
> > Again I'm confused. Presuming the signal is in the 15 Mhz range (67 ns perio
d)
> > and a PIC is running with a 5Mhz instruction clock (200 ns period), then eve
n
> > if such a chip existed, the finest resolution that the PIC could control the
> > counter is at 3 count intervals, which is the number of ticks the counter
> > would clock inbetween instructions. Right?
> >
> > Is the exact count necessary? Then you need something quite more complex tha
t
> > could be programmed with the amount of time to gate the incoming pulses alon
g
> > because the PIC isn't fast enough.
> >
> > If an exact count isn't necessary, then you simply prescale the count and le
t
> > the PIC count it.
> >
> > But in any case the PIC's prescaler and count hardware is fast enough to
> > handle the signal....
> >
> > BAJ
>
> Well, its application requires a minimum counting of 15MHz, and it is
> not
> a pulse counter, but a frequency counter, what means it can stay holding
> the counting as long it needs to read the shift register.

So what's the maximum frequency?

>
> The counting gate is easily calculated with the PIC instruction
> execution
> time versus quantity of instructions, the common time loop, so it is
> easy
> to make a high precise counting gate even with a pic. Using a round
> number
> crystal frequency it will go to the crystal precision and accuracy.

But if the frequency of what is counted is a lot higher than the frequency of
the PIC then you cannot have precise control over the precision of the counts.
No to within one pulse. Say you have a frequency that's 8 times faster than
the PIC. The accuracy of the count you get will only be within 8 pulses of
the actual number of pulses due to the slowness of the PIC's clock. So would it
then not be better to simply prescale the pulses by 8 and let the PIC count
the results?

>
> There is another problem involved, I didn't say at the first post, the
> pic
> will serve just as an interface between the device that will make use of
> the data and the 8 (yes, eight) counters like that working at the same
> time.

I guess the real question is there a requirement for each of the counts to
be taken at the same time. Since the interface PIC can hold and count as
necessary, simultaneous counting doesn't seem to be a strong requirement. If
that's the case then a simple multiplexer should be sufficient to gate the
appropriate signal to the PIC.

>
> The interesting point in here is that a simply 8 jk's fflops counter and
> an
> incorporated shift register could not cost more than $0.60, while it
> would
> be just the PIC crystal's cost.

But the part doesn't exist AFAIK. You'd have to have 2 parts, a counter like
a 74HC393 and a loadable shift register. So to get simultaneous counting
you'd need at least 16 parts. Whereas if you use the PIC, which is already
there. You only need to add a single multiplexer.

Can you be more specific about your design requirements? What are the min
and max frequency of the sampled signals. Do they need to be sampled
simultaneously?

BAJ