Measurement Computing WBK17 - Manual (Page 15)

WBK17, Counter/Encoder Module 987996 WBK17, pg. 15
PERIOD: OPT6: This allows a mapped channel’s period to be measured, instead of the input channel.
The mapped channel can be any other input channel (post debounce) or one of the detection signals (on any
channel). This option allows the counter to be used with any other input channel (post-debounce), or with
one of the detection signals. If the channel’s input is used elsewhere, for example, gating another counter,
the counter for this channel does not need to go unused.
Period and Frequency Accuracy
The WBK17 can measure the period of any input waveform. It does this by counting the integral number of
“ticks” that make up the period, the data returned will always be time measured in “ticks.” The error in
each data sample will come from two sources: the sampling error caused by not being able to count a
partial “tick”; and the WBK17’s internal timebase inaccuracy. The WBK17’s internal timebase has an
absolute accuracy of 10ppm. The sampling error will vary with input frequency, selected ticksize, and
selected averaging mode. The absolute error is the “root-sum-of-squares” of the two independent error
sources. For example, if the sampling error is 10ppm and the timebase accuracy is 10ppm, the absolute
accuracy is 14.1 ppm.
Many times the desired accuracy is much less than what the internal timebase is capable of. Other
applications will require a more accurate period measurement and the effects of sampling error will have to
be averaged out leaving only the inaccuracy associated with the internal timebase. Inaccuracy due to the
internal timebase cannot be averaged out.
For period and frequency measurements, percent sampling error is equal to 100%/(n+1) where n=0 to
65,535 for a 16-bit counter and n=0 to 4,294,967,295 for a 32-bit counter. For small count values, the
sampling error is large and for large count values, the sampling error is small. If sampling error is to be less
than 0.2%, n must be greater than 499 regardless of counter size. If sampling error is to be less than
0.001% (10 ppm), n must be greater than 99,999 which requires use of a 32-bit counter. These values of n
were used to derive the frequency range tables below.
Sampling error can also be reduced by averaging many samples together. Assuming the input signal is
asynchronous to the WBK17’s internal timebase, sampling error can be divided by the square-root of the
number of samples taken. The averaging can be done with PC-based software.
The WBK17 has the ability to measure 1, 10, 100 or 1000 periods, dividing the sampling error by 1, 10,
100, or 1000. This is done inside the WBK17 and may eliminate the need for any averaging to be done in
the PC. For high accuracy on high frequency inputs, multiple period measurement and PC-based averaging
can be done.
The WBK17 has the ability to provide various frequency ranges that are based upon different ticksizes,
averaging options, and counter size (16 bit or 32 bit values.) The frequency ranges are designed to fit a
wide array of possible applications. Within each range, the sampling error decreases dramatically as the
input period increases. The ranges will get smaller as required accuracy increases.
Upper 16-bits of the 32-bit counter Lower 16-bits of the 32-bit counter
Range (Hz) Ticksize
(nS) Averaging
Option
Range (Hz) Ticksize (nS) Averaging
Option
15u – 1500u 20000 1 1 – 100 20000 1
150u – 15m 2000 1 10 – 1k 2000 1
1500u – 150m 200 1 100 – 10k 200 1
15m – 1500m 20 1 1k – 100k 20 1
150m – 15 20 10 10k – 1M 20 10
1500m – 150 20 100 100k – 5M 20 100
15 – 1500 20 1000 1M – 5M 20 1000
Frequency Ranges for a 16-bit value, sampling error is less than 0.2%