Announcement

Collapse
No announcement yet.

OT - how does a frequency counter typically sample?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • OT - how does a frequency counter typically sample?

    If I have a stead stream of electrical pulses exactly say 1/1000 of second apart, I'd expect a frequency counter to tell me the signal is 1000 Hz or 1 kHz.

    But what if the signal if a repeating pattern of one .0005 sec pulses follow by one .0015 seconds.....you still get 1000 pulses per second but each pulse is quite a different length.

    Does the counter say sorry I can't read this or does it tell you a frequency of 1 kHz?

    thanks
    .

  • #2
    I think most have an input that will be rated to some kind of "duty cycle". Like 0-20kHz 10-90% duty cycle. The duty cycle is the percentage the signal is in its HIGH state vs. LOW state.

    The counting method probably detects the rising or falling edge of the signal and then increments a counter, and after a 1 second timer has elapsed, the unit displays the count and resets to 0. The detect and increment circuits take time to work, so if the frequency is too fast it may not be able to detect. Short pulses with long gaps between them should be alright, I'd expect.

    Issues with capacitance on the meter input side can effect the minimum duration of a spike that can be detected.

    EDIT

    Regarding your last question; the counter doesn't know what it misses. It will only count pulses it can recognize and anything too short or above the frequency limit is not counted.

    Comment


    • #3
      Originally posted by Mcgyver View Post
      If I have a stead stream of electrical pulses exactly say 1/1000 of second apart, I'd expect a frequency counter to tell me the signal is 1000 Hz or 1 kHz.

      But what if the signal if a repeating pattern of one .0005 sec pulses follow by one .0015 seconds.....you still get 1000 pulses per second but each pulse is quite a different length.

      Does the counter say sorry I can't read this or does it tell you a frequency of 1 kHz?

      thanks
      It should still read 1 kHz, provided the pulses are long enough for the counter's time base. I have built custom frequency counters and I have several lab grade frequency counters on my bench. I use them to count pulses from scintillators. The pulses range from less than 1 microsecond to several microseconds in duration, but the counter doesn't care. I can set a counter to sample either the rising edge or falling edge at a particular level. That's all it cares about, not what happens after the edge.

      Comment


      • #4
        You know... I don't know. On my o-scope it would show an irregular trace or a fuzzy trace depending, but on the frequency counter on my multi meter it would probably just read wrong. I once tested the output for a generator and my meter said 400hz, but a 3600 rpm motor plugged in turned 3600 rpm, so obviously the meter was confused by some unfiltered noise. The same meter on my household electrical read 60hz like you would expect.
        *** I always wanted a welding stinger that looked like the north end of a south bound chicken. Often my welds look like somebody pointed the wrong end of a chicken at the joint and squeezed until something came out. Might as well look the part.

        Comment


        • #5
          On the very simple frequency counters I have designed and built the count of rising (or falling) edges that occurs within the count window is the determining factor. With the variable duty cycle described in the O.P., framing errors are induced and will cause at least a +/-1 count or greater variation of successive counts to occur, causing jitter in the resulting count display. More sophisticated counters will perform and display an average of successive count windows.
          Weston Bye - Author, The Mechatronist column, Digital Machinist magazine
          ~Practitioner of the Electromechanical Arts~

          Comment


          • #6
            All the frequency counters I've used during my working career counted transition pairs past a given threshold voltage that you could set. Each excursion pair being an upward crossing then a downward crossing. So duty cycle did not enter into the measurement. If you got 2000 threshold crossings it read out as 1kHz. Duty cycle for each pulse did not enter into the counting.

            Comment


            • #7
              Thanks for the info gents. In case it wasn't clear, it wasn't really a change in dutycyle, it was a change in the period of a complete cycle.

              the application is a watch hairspring vibrating tool. Its a particularly challenging aspect of watchmaking - determining the exact length of the hairspring (which determines the oscillation rate). You determine the length by trial and error, holding the spring at different spots until you get the frequency you want.

              You set the balance oscillating and using the arms of the balance to break a light beam to generate a count. Here's a video of a commercial one that makes its function clear. They're about 3000 Euros so i will have to roll my own. I've a HP 5225a not working....assuming I can make it work I can use it to count the number of times an arm of the balance wheel breaks the light.....but it would be almost impossible to get the sensor in exactly the middle of the arms arc of travel (which it would have to be for the pulses to be the same length).....so one pulse, complete cycle from say trigger of the signal going high back to going high again won't be the same length as the next and that pattern will alternate.
              Last edited by Mcgyver; 12-21-2017, 12:51 PM.
              .

              Comment


              • #8
                Originally posted by Mcgyver View Post
                But what if the signal if a repeating pattern of one .0005 sec pulses follow by one .0015 seconds.....you still get 1000 pulses per second but each pulse is quite a different length.
                Actually, with those numbers the period equals 0.002 seconds resulting in 500 pulse per second. To address your question, the signal being
                measured does not need to have a 50% duty cycle. The counter may specify a minimum pulse width requirement in order to be able to
                reliably detect an edge, but that number would tend to be small. The counter always counts on the same edge of the waveform so duty cycle
                is irrelevant.

                Comment


                • #9
                  Originally posted by Mcgyver View Post
                  Thanks for the info gents. In case it wasn't clear, it wasn't really a change in dutycyle, it was a change in the period of a complete cycle.

                  the application is a watch hairspring vibrating tool. Its a particularly challenging aspect of watchmaking - determining the exact length of the hairspring (which determines the oscillation rate). You determine the length by trial and error, holding the spring at different spots until you get the frequency you want. ...
                  The counter just counts pulses during a window in time. If the frequency is changed during that window, you'll have to wait for the next window
                  to get an accurate measurement.

                  Comment


                  • #10
                    Originally posted by Mcgyver View Post
                    Thanks for the info gents. In case it wasn't clear, it wasn't really a change in dutycyle, it was a change in the period of a complete cycle.
                    Generally, that won't matter. It just counts edges - in my applications, pulses can come at virtually anytime (random radioactive decay) and I can choose the time window from infinity (simple counter) down to whatever the lower limit of the time base is. Suppose the window is 1 second and you have 1000 pulses - the counter doesn't care if those all arrive in the first 100 milliseconds or equally through out the 1 second window (assuming that it's sampling fast enough to catch all the transitions in that first 100 milliseconds, of course). It will still report is as 1 kHz. If the window and display update is smaller than 1 second, you could see some flicker on the display (e.g. 1 kHz followed by 0 Hz).

                    Edit: When I say "that won't matter", I was thinking in terms of my applications where I care about total counts or count rates, NOT accurate frequency measurements. As RichR points out, to get an accurate measurement of the frequency, the pulses need to be stable during the window.

                    Comment


                    • #11
                      ok, I think it get it.....they will be very stable, but alternating....ie pattern of the periods might .21000 .19000 .21000 .19000 so from this, I'd expect a reading 5 Hz, assuming the window was say a second (perhaps it can be set). The error might be if the counter got 3 .21000 and 2 .190000 vs an even number of both within the window, correct?
                      .

                      Comment


                      • #12
                        Originally posted by Mcgyver View Post
                        ok, I think it get it.....they will be very stable, but alternating....ie pattern of the periods might .21000 .19000 .21000 .19000 so from this, I'd expect a reading 5 Hz, assuming the window was say a second (perhaps it can be set). The error might be if the counter got 3 .21000 and 2 .190000 vs an even number of both within the window, correct?
                        Thats correct, it will basically average any jitter. A lot of counters can also measure period directly. Depending on the frequency of interest, that
                        could provide faster measurements and higher resolution. Google doesn't return anything for HP 5225A, are you sure about that number?
                        Last edited by RichR; 12-21-2017, 01:49 PM. Reason: Grammer

                        Comment


                        • #13
                          Originally posted by RichR View Post
                          Google doesn't return anything for HP 5225A, are sure about that number?
                          sorry 5335a
                          .

                          Comment


                          • #14
                            A lot of people assume that all frequency counters operate as "frequency counters*" in the sense that they count the number of zero crossings during a fixed gate time. The counter you have (5335a) has the architecture of a time-interval counter, so it may or may not not behave in the way you expect if the input consists of something other than a stable sine wave, pulse train, etc. In your case, you might be better off with a simple "frequency counter."

                            *In addition to frequency counters, there are reciprocal counters, time interval counters, harmonic heterodyne counters, etc. as well as counters that can take on multiple measurement personalities.

                            Comment


                            • #15
                              If the frequency is stable, you can use a phase locked loop and a divide-by-N counter to get higher precision. I once designed a power line frequency meter that provided a signal of 60 kHz from a 60 Hz source, so it was capable of reading 60.000 Hz with a one second update time. I used a 74HC4046 PLL and VCO with a couple of 74HC393 dual decade counters.

                              https://en.wikipedia.org/wiki/Phase-locked_loop
                              https://en.wikipedia.org/wiki/Frequency_multiplier
                              https://www.electronics-tutorials.co...s/74hc4046.htm
                              http://www.ti.com/lit/ds/symlink/cd54hc393.pdf
                              http://pauleschoen.com/pix/PM08_P76_P54.png
                              Paul , P S Technology, Inc. and MrTibbs
                              USA Maryland 21030

                              Comment

                              Working...
                              X