[sdiy] Re: S[DIY} frequency counter

[Magnus Danielson]

David G. Dixon wrote:
>> It depends on your goal. If you're trying to learn something about
>> digital design, and this group is about doing it yourself, then using a
>> few (obsolete?, unavailable?) chips that do everything for you and
>> wiring it up per a datasheet is fairly pointless.  You may as well go
>> buy a cheap multimeter with built-in freq counter, gut it out of its
>> case, and stick it on a panel.
> 
> More importantly, it depends on what you actually want to achieve with the
> device.  I would like to be able to measure fractions of Hz at the low end
> of the audio spectrum or even below.  Hence, I will probably end up adding a
> PLL to multiply the signal...  

Adding a PLL for that range would be both more trouble than you want and 
totally pointless from a counting precission perspective.

Consider that you want to measure a 1 Hz signal, consider that your 
timebase is a 10 MHz crystal oscillator. Then by measuring one cycle you 
have 7 digits of resolution right there. The lower frequency, the higher 
resolution. Of course your accuracty and precission wont follow along, 
in particular the trigger curcuit and the time-base will prohibit that.

Counters can be divided into three groups:

Fixed time: Count the number of input cycles during the fixed time. As 
frequency increases the resolution increases. Resolution can be 
increased by making the gate-time longer but this lowers the readout speed.

Fixed event count: Counts the number of reference cycles during a fixed 
number of input cycles (number of events). As the frequency decreases 
the resolution increases. Resolution can be increase by making 
event-count larger but this lowers the readout speed.

Both these two types have increased resolution in one or the other end 
of the scale, so they where bound to be combined, so therefore:

Reciprocal counter: Counts the reference cycles and input cycles (event) 
during the gate time. The resolution increases both at the high and low 
frequencies. Resolution increases as gate time is increased.

For all these counters, readout can be made by calculating:

Freq = Event/Time

Period = Time/Event

Where time is the counts of reference oscillator time the period of the 
reference oscillator.

All these types can have improved by using interpolator curcuits that 
improve the time-resolution at the start and stop of the gating period. 
A virtually higher reference oscillator is produced. For instance, in 
the HP 5335A the analog interpolators provide 200 time improved gain, so 
the 10 MHz time base oscillator achieves 1 ns of resolution (you have a 
+/- 1 cycle problem which removes some of the resolution).

An typical chip-based counter like the Intersil is of the first or 
second kind depending on weither it measures frequency or period.

>The Art of Electronics has a very nice
> subchapter on frequency counting and period measurement, by the way.

It's not bad for such a wide-range book.

Cheers,
Magnus